AACE International Annual Meeting 2020

Last month marked my second attendance at an Annual meeting of AACE International (formally Association for the Advancement of Cost Engineering).  The Covid-19 Pandemic forced a change of venue from Chicago to Virtual/Online, with overall good results.  Most of the technical presentations were pre-recorded for viewing at leisure.  There was also a live track including keynotes, selected technical presentations, and business meetings.  There were attempts to facilitate online “networking” and other informal “events”, but I missed those.

Earlier in the year, I co-authored a paper (with Patrick M. Kelly) for presentation at the meeting: (PS-3496) Interpreting Logic Paths in Multi-Calendar Project Schedules.  It’s essentially a deep-dive into the Multiple Float Path calculation in Oracle Primavera P6 project scheduling software, expanding on (and in a few cases contradicting) my earlier blog articles on the subject.  Pre-recording the presentation (as a narrated PowerPoint show) involved quite a bit more effort than a live presentation would have; I hope to become more efficient at it with additional practice.

The presentation (all 70+ minutes of it) and the paper are included in the Transactions bundle for registered attendees of the meeting.  They are also available for purchase by non-attendees (including a training certificate after viewing) at the AACE store.  I think the price is $25 for members of AACE and $30 for non-members.  Eventually I’ll look to host a form of the paper on the Links page of our own website.

I also got “elected” to the board of the Planning and Scheduling Subcommittee, and I look forward to taking over the coordination of our Recommended Practices from our rising Chairman, Jessica Colbert, for the next year.

Driving Logic Relationships in Project Schedules

[Article 1 of 2.] This is a summary of the standard definitions and uses of driving logic relationships between activities in project schedules, as applied in Primavera P6 and Microsoft Project software.  Driving relationships are often considered fundamental elements of the project critical path.

This winter I worked with a colleague to prepare a paper – Interpreting Logic Paths in Multi-Calendar Project Schedules – for presentation at this year’s AACE International Conference and Expo in Chicago (Covid-19) virtual world.  The paper reflects a deep dive into the Multiple Float Path calculation options in Primavera P6 scheduling software.  During the technical study, I had a lot of opportunities to think about driving logic relationships.  This entry summarizes the standard definitions and uses.  I’ve summarized a couple alternate definitions and uses in another article.

The Importance of Driving Logic

The planning and execution of complex projects requires the project team to understand, implement, and communicate the consequences of schedule logic flow to the other stakeholders.  Through schedule logic, each activity in the project has the potential to constrain or disrupt numerous other activities – and to be constrained or disrupted by them.  The most obvious artifacts of logic flow are the important logic paths, like the critical path, the Longest Path (in Primavera P6), or the driving path to a key delivery milestone.  Regardless of the detailed definition, each of these important paths is governed by driving logic relationships from the first activity to the last activity in the path.

Standard Definition and Uses of Driving Logic Relationships

A driving relationship is “A relationship between two activities in which the start or completion of the predecessor activity determines the early dates for the successor activity with multiple predecessors.  See also: Free Float.” [That’s the standard definition from AACE International.]  Alternately, “A driving relationship is one that controls the start or finish of a successor activity.” [That’s from the PMI publication on CPM Scheduling for Construction.]

For practical purposes, a driving relationship is a predecessor relationship that prevents a successor activity’s early start or early finish from being scheduled any earlier than it is.  When an otherwise unconstrained activity has only one predecessor, then it is normally, and obviously, a driving predecessor.  When an activity has multiple predecessors, then one or more of them may be driving while the others are non-driving.  These distinctions answer the key questions, “Why is this activity scheduled when it is?  Why can’t we do it sooner?”

Driving Logic in Primavera P6

Like its predecessors, P6 routinely illustrates driving logic relationships using solid lines on bar charts – either red or black depending on the “Critical” status of the connected activities.  Non-driving relationships are depicted with the same colors, and those that are also non-critical use dotted lines.  This is demonstrated in the figure below, where the non-critical activity “Task” (A1020) has two predecessors and five successors. One of the predecessor relationships and all five of the successor relationships are depicted with solid black lines and marked as driving (but not critical) in the relationship tables.  The non-driving relationships – one from “pred a” to “Task” and five more from Task’s successors to the project’s finish milestone – are depicted with dotted lines.  The two critical, driving relationships that connect the “CP” activity to the project’s start and finish milestones are depicted with solid red lines.

In small projects it is often easy to identify driving logic flow in printed P6 bar charts by visually tracing the solid relationship lines between activities.  As project schedules become larger and more complex, however, the number of relationship lines increases to the point that visual tracing becomes impractical.  Then driving relationships are primarily identified using the relevant columns of the associated relationship tables.  Experienced P6 users often use the “GoTo” buttons in the relationship tables to click-trace along driving logic paths – backward and forward through complex project schedules – to review and confirm important chains of sequential logic (i.e. driving logic paths).

In general, Primavera P6 identifies driving relationships by analyzing the intervals between early dates of the linked activities, after completion of the core scheduling calculations.  With a few minor exceptions, a driving relationship is identified when the Relationship Successor Free Float (RSFF) equals zero.  In addition to providing a basis for the graphical and tabular depictions of driving logic flow, P6 uses these driving attributes to automatically identify the Longest Path, or the driving path to project completion.

Driving Logic in Microsoft Project

Unlike P6, Microsoft Project (MSP) does not graphically differentiate driving and non-driving relationship lines in Gantt-chart views, and the standard relationship (i.e. dependency) tables provide no driving-logic indicators.  The Task Inspector pane provides the primary method for identifying driving predecessors of the currently-selected task; there is no corresponding method for identifying driven successors.  The figure below depicts the same schedule as before, now in MSP format, with the Inspector pane identifying a single (driving) predecessor task, “pred b”, for the currently-selected task (“Task.”)  As far as it goes, this agrees with P6.

Although not presented to users, driving relationship indicators are developed by MSP (at least since MSP 2007), with the results being stored in the PredecessorDrivers collection for each task.  This collection forms the basis of the Predecessors list of the Inspector pane.

It’s also apparent that the PredecessorDrivers data are used to define the Driving Predecessors and Driven Successors bar styles that were introduced as part of the Task Path functionality in MSP 2013.  This functionality is illustrated below, where the driving and non-driving predecessors of “Task” – and its driven successors – are all differentiated by bar color.  Although there are clear limitations to this graphical approach, the ability to show driving and driven logic paths (if not individual driving relationships) is a major improvement for MSP users.

Unfortunately, the internal MSP calculation of driving logic attributes – and the explicit paths of driving logic that they purport to illustrate – have proven unreliable for complex schedules with other than finish-to-start relationships, out-of-sequence progress, or external links.

Standard Driving Logic in BPC Logic Filter for Microsoft Project

BPC Logic Filter (my company’s add-in for MSP) identifies driving logic by independently analyzing relationship relative floats after completion of the schedule calculations.  This is a bit like the P6 approach and has proven, at least for me, more reliable than the internal MSP data when things get complicated.  This figure shows the combination of a logic tracer view (with special bar styles depicting driving and near-driving logic paths) together with the task logic inspector tables.  Driving relationships are highlighted yellow in the tables.  Overall, this seems to combine the best parts of the corresponding P6 and MSP layouts, and the Jump buttons allow for logic-based navigation forward or backward through the schedule network.

Continue to related article…

Alternate Definitions of Driving Logic Relationships in Project Schedules

Construction CPM Conference 2020 – Avoid Out of Sequence Progress in Microsoft Project

Click Here to view the screen and audio capture of my presentation at the 2020 Construction CPM Conference.  Fred Plotnick (the Conference organizer) is kind enough to host the content on his server.

The presentation recaps and expands on my blog entry on the same subject: Avoid Out-of-Sequence Progress in Microsoft Project 2010-2016.

Overlapping Tasks in Project Schedules

In a project schedule, overlapping tasks are tasks that are BOTH sequential and concurrent.  The effective and efficient scheduling of overlapping tasks typically requires the use of time lags and logic relationships that are not Finish-to-Start.

Introduction

In a project schedule, overlapping tasks are tasks that are BOTH sequential and concurrent, with neither condition being absolute.  They can exist in highly-detailed schedules for project engineering, design, production, and construction, though they are much more common in high-level summary schedules.  Essentially, Task A and Task B are overlapping when:

  1. Task B is a logical successor of Task A; and
  2. Task B can start before Task A is finished.

Handling the second condition typically requires the use of time lags and logic relationships that are not Finish-to-Start.  These items were not supported in the original Critical Path Method (CPM), and some scheduling guidelines and specifications still prohibit or discourage their use.  Nevertheless, they remain the most effective tools for accurately modeling the plan of work in many cases.

There are essentially two categories of overlapping task relationships: Finish-Start Transitions and Progressive-Feeds.

Finish to Start Transitions

Overlapping tasks with Finish-Start Transition relationships can be described in terms of a relay (foot)race, where an exchange zone for handing over the baton exists between each pair of “legs” (i.e. the 4 stages of the race).  At the end of Leg 1, Runner 2 must start running before Runner 1 arrives, timing his acceleration to ensure an in-stride passing of the baton in the exchange zone, simultaneous with the completion of Leg 1.  Runner 2 does not care about Runner 1’s fast start nor his awkward stumble at the midway point; his only focus is on gauging Runner 1’s finishing speed and starting his own run at the precise instant necessary to match speeds in the transition zone.  In practice, Runner 2 establishes a mark on the track – paced backward from the exchange zone – and starts his own run when Runner 1 reaches the mark.

Real-world examples of such overlap include the cleanup/de-mob and mob/setup stages of sequential tasks in construction projects.  In engineering/design, many follow-on tasks may be allowed to proceed after key design attributes are “frozen” at some point near the finish of the predecessor task.  In general, the possibility of modest overlap exists at many Finish-Start relationships in detailed project schedules, sometimes being implemented as part of a fast-tracking exercise.  The most common front-end planning occurrence of these relationship in my experience is in logic driven summary schedules, where analysis of the underlying detailed logic indicates that the start of a successor summary activity is closely associated with, but before, the approaching finish of its summary predecessor.

In terms of a project schedule, this kind of relationship is most easily modeled as Finish-to-Start with a negative lag (aka “lead”).  This is illustrated by the simple project below, where tasks A, B, and C are sequentially performed between Start and Finish milestones.  Each task has a duration of 9 days to complete 90 production-units of work.  (A linear production model is shown for simplicity.)  Because of the Finish-Start Transition, Task B and Task C are allowed to start 1 day before their predecessor finishes. 

Negative lags can be used for date manipulation, such as to hide an apparent delay.  Negative lags can also be the source of float (and critical path) complications for project schedules with updated progress, particularly when the lag spans the data date.  Consequently, negative lags are discouraged or explicitly prohibited in many schedule standards and specifications.

When negative lags are prohibited, overlapping tasks with Finish-Start transitions may be modeled by breaking the tasks into smaller, more detailed ones – all connected with simple Finish-Start links and no lags.  In the example, the two negative-lag relationships can be replaced by two pairs of concurrent 1-day tasks – the last part of the predecessor and the first part of the successor – that are integrated with FS links.  Thus, the overlapping linear production of the three tasks now requires 7 tasks and 8 relationships to model, rather than the original 3 tasks and 2 relationships.  Alternately, each pair of concurrent tasks could be combined into a single “Transition” task, though such an approach could involve additional complication if resource loading is required.

In practice, the extra detail seems hardly worth the trouble for most schedulers, so simply ignoring the overlap seems fairly common.  This has the consequence of extending the schedule.

By ignoring the overlap, the scheduler here has added two days (of padding/buffer/contingency) to his overall schedule, extending the duration from 25 to 27 days.  This is unlikely to be recovered.

Overlapping Tasks with Progressive Feeds

The predominant category of overlapping tasks involves repetition of sequentially-related activities over a large area, distance, or other normalized unit of production.  The activities proceed largely in parallel, with the sequential relationships based on progressive feeding of workfront access or work-in-process units from predecessor to successor.  In construction, a simple example might include digging 1,000 meters of trench, laying 1,000 meters of pipe in the trench, and covering the trench.  The most timely and profitable approach to the work is to execute the three tasks in parallel while providing adequate work space between the three crews whose production rates are well matched.  This is often described as a “linear scheduling” problem; common examples in construction are railways, roadways, pipelines, wharves, industrial facilities, and even buildings (e.g. office towers – where steel, concrete, mechanical, plumbing, electrical, finishing, and trim activities need to be repeated for each floor.)  Many large-scale production/manufacturing operations are set up to maximize overall throughput by optimizing the progressive feeding of production units through the various value-adding activities.  Proper scheduling of such activities is necessary when similar techniques are applied in non-manufacturing industries like construction, e.g. production lines for precast concrete piles or panels.

Below is a simple table and associated linear production chart summarizing three sequential tasks (A, B, and C) that must be repeated 90 times along a workfront to complete a specified phase of work.  Each task can be executed 10 times per day, resulting in a 9-day duration for the required 90 units of production.  For safety and productivity reasons, it is necessary to maintain a minimum physical separation of 30 units (i.e. 3 days’ work) between the tasks at all times.  Thus, Task B must not be allowed to start until Task A has completed 30 units of production (~3 days after starting), and it must not be allowed to complete more than 60 units of production (~3 days from finishing) until Task A has finished.  Task C must be similarly restrained with respect to Task B.  As a result, the overall duration of the three tasks is 15 days.

The three tasks must now be incorporated into a logic-based project schedule model.  When doing so, the following potential issues should be kept in mind:

  • In most scheduling tools, relationship lags are based on an implied equivalence between production volume (or workfront advancement) and time spent on the task. The validity of the lags needs to be confirmed at each schedule revision or progress update.  (One exception, Spider Project, may offer more valid methods.)
  • Using progressive-feed assumptions with unbalanced production rates can have unintended consequences. For example, if the production rate of Task B is doubled such that the task can be completed in half the time, then the start of the task may be delayed to meet the finish restraint.  This is consistent with a line-of-balance planning philosophy that places the highest priority on the efficient use of resources, such that scarce or expensive resources will not be deployed until there is some assurance that the work may proceed from start to finish at the optimum production rate, without interruption.  In the example, the delayed start of Task B also delays the start of Task C, leading to an increase in the overall project duration from 15 days to 20 days.  Some writers refer to this phenomenon as “Lag Drag.”  The overall schedule is optimized when progressive-feed tasks are managed to the same balanced production rate, and disruptions are minimized.
  • A progressive-feed model may not be valid if the physical or temporal requirements underlying the lags at task Start and Finish are violated during task execution. For example, if the daily production rate of Task A follows a classic S-curve profile (“Task A Logistic”) while Task B’s stays linear, then maintaining the required 30-unit minimum physical separation may require additional delay at the start of the second task.

Compound Relationships: The Typical Approach in Oracle Primavera P6

As shown in the following figure, scheduling these overlapping tasks is fairly straightforward in P6.  Because P6 supports multiple relationships between a single pair of tasks, it is possible to implement the required Start and Finish Separations as combined Start-Start and Finish-Finish relationships, each with a 3-day lag.  These are also called “Compound Relationships.”  The resulting representation of the linear schedule is completed with 3 tasks and 4 relationships (excluding the Start and Finish milestones.)  The three tasks are likely well aligned with the labor and cost estimates for the project, so resource and cost loading of the schedule should be straightforward.  The scheduler must still ensure that the three concerns above are addressed, namely: validating lag equivalence to work volumes or workfront advancement, balancing of production rates, and confirming lag adequacy when used with differing task production profiles.

One-sided Relationships: The Typical Approach in MSP

Microsoft Project does not permit more than one relationship between any two tasks in a project schedule (Ladder Logic in Microsoft Project).  As a result, the scheduler in MSP will typically choose to implement either a Start-to-Start or Finish-to-Finish restraint with a corresponding lag.  Both options are shown in the following figure.

In either case, the resulting schedule will have the lowest number of tasks {3} and relationships {2} to manage (for both cost and schedule) through the project.  This approach is easy to implement.

The most obvious problem with this typical approach is the inadequate logic associated with the dangling starts and dangling finishes (Dangling Logic).  As a result, the typical CPM metrics of Slack (i.e. Float) and identification of the Critical Path will not be reliable, especially after the start of progress updates.

PDM with Dummy Start Milestones

Correcting the dangling logic issues in MSP schedules is most simply addressed using dummy milestones to carry either the start or finish side of the logic flow.  Below I’ve shown two variations using dummy Start milestones:

Alternate A involves Trailing Start Milestones.  Here, the milestones exist as Start-to-Start successors of the corresponding tasks, effectively inheriting their dates from the corresponding task Start dates.  The trailing start milestones pass logic to the successor tasks via relationships of the form, Start-to-Start-plus-lag.

Alternate B involves Leading Start Milestones.  Here the milestones exist as Start-to-Start successors of the preceding tasks (plus lags) and as Start-to-Start predecessors of their corresponding tasks (no lags).

The two alternates are largely equivalent, though Alternate B (Leading Start Milestones) has one significant advantage: it works with percentage lags.  When a percentage lag is imposed, the imposed time lag increases or decreases as the predecessor’s duration increases or decreases.  This reduces some of the risks of the assumed production volume = time equivalence.  (Be careful, though; the imposed lag is always a percentage of the overall Duration of the predecessor task, having nothing to do with the Actual (i.e. to-date) Duration.  Moreover, all lags in MSP are imposed using the successor-task’s calendar, so mis-matched predecessor and successor calendars can bring surprises.)

Using the dummy milestones presents valid schedule logic with a relatively modest addition of detail (i.e. medium number of tasks {5} and relationships {6}.)  The schedule stays fully aligned with labor/cost estimates; no deconstruction is required, and it responds well to unbalanced and varying production rates.  Unfortunately, the dummy milestones can cause visual clutter, so presentation layouts need filters to remove them from view.

Full-Detail: the CPM Ideal

Non-Finish-to-Start relationships were not supported in the original CPM, and they are discouraged or prohibited in some scheduling standards and specifications.

If only Finish-to-Start relationships are allowed, then accurate modeling of the three overlapping tasks requires substantial deconstruction into a larger number of detailed subtasks.  For the three-task example, the schedule model below breaks each 9-day task into nine 1-day tasks, all integrated with Finish-to-Start relationships.  The model is depicted using MSP; a similar model could be constructed in P6.

Overall, this approach appears to be more “valid” with respect to pure schedule logic.  That is, there are no leads, no lags, and no non-Finish-to-Start relationships.  The resulting model can also respond well to unbalanced and varying production rates, and it is likely to stay valid through progress updates.

On the “con” side, this model has the maximum level of detail (i.e. highest number of tasks {27} and relationships {39}.)  Consequently, it will introduce substantial complications to resource and cost loading, and it will be the hardest to manage through completion.  More importantly, the logic relationships that accompany such additional detail are not always technologically required.  While the ordering of Units 21-30 prior to Units 31-40 may appear perfectly reasonable in the office, all that really matters is that ten units of production are received, completed, and passed on to the next task each day.  The addition of such (essentially) preferential logic increases the chances that the actual work deviates substantially from the plan, as field conditions may dictate.  That can severely complicate the updating of the schedule, with no corresponding value added.

Compromise: PDM with Partial Detail

The next figure presents a compromise, providing additional task details as needed to address the initial separation requirements but minimizing the use of lags and non-Finish-Start relationships.  The result is a moderate schedule with “mostly valid” logic and only modest level of detail (i.e. medium number of tasks {6} and relationships {7}.)  Such a schedule presents medium difficulty of implementation and is less susceptible to the “preferred logic” traps identified earlier.  It also responds well to unbalanced and varying production rates, and it stays valid (mostly) during progress updating.

This schedule still requires consideration and validation of the progressive-feed assumptions.  Since this schedule is only partly aligned with existing labor/cost estimates, some de-construction of those estimates may be required for resource and cost loading.

Resource Leveling Changes from MSP 2010 to MSP 2016 – Revisited

In a departure from an earlier study, the resource-leveled schedule generated using Microsoft Project (MSP) 2016 may sometimes be substantially shorter than the comparable schedule from MSP 2010.  The result suggests a more sophisticated leveling algorithm that, while sometimes generating shorter schedules, may increase the associated resource-related risks.

Several years ago I wrote an article examining some reported differences in resource leveling behavior between Microsoft Project (MSP) 2010 and 2013 versions.  In general, those observations implied that the newer software was generating longer schedules under default conditions.  (The observations were by others, since at the time I was only using MSP 2010.)  I concluded the article with a speculation that the leveling algorithm may have been adjusted to preserve the appearance of the pre-leveling “Critical Path,” but at the expense of a longer schedule.

A recent test case using MSP 2016 does nothing to confirm that speculation, and an opposite conclusion is implied.

Below is an idealized schedule for construction and commissioning of a small processing plant, including resource loading.  Only technologically-required logic is included, so that a) the resources are all severely overallocated, and b) the schedule is unrealistic.  The same schedule is shown in both MSP 2010 and MSP 2016 forms, with only cosmetic differences.  There is a Deadline of 25Feb’05 on the Substantial Completion task, and no explicit risk buffers are present.  [The MSP 2010 form was used to illustrate the resource-leveled critical path in a technical paper I presented at AACE International in 2017.] 

To resolve the resource over-allocations and generate a more realistic schedule, the resource leveler is applied in both MSP 2010 and MSP 2016 using simple leveling options with no task priorities defined.

As shown below, these options lead to completely different schedules in the two tools, with MSP 2016 generating a leveled schedule that is 10-days shorter than the MSP 2010 version.

Key observations:

  1. The MSP 2010 schedule failed to meet the Deadline, so more tasks are marked Critical due to zero-or-negative Total Slack. Six tasks including the two milestones have TS=-10d (i.e. the “most critical” as defined by Total Slack.)
  2. The MSP 2010 leveler did not impose any task splits, even though they were allowed.
  3. The MSP 2016 schedule meets the Deadline, so no negative slack is imposed. Seven tasks including the two milestones have TS=0 (the “most critical.”)
  4. The MSP 2016 leveler imposed a task split on one task (A3 Structures).

A close look at the Resource-leveled Critical and Near Critical paths – Using the Near Longest Path Filter in our BPC Logic Filter add-in – demonstrates that MSP 2016’s leveler creates a more condensed schedule.  Specifically:

  1. MSP 2010 generates a single continuous Resource-leveled Longest Path (BPC Relative Float = 0) comprised of eight tasks in sequence. The other 12 tasks possess 10 to 70 days of relative float.
  2. MSP 2016 generates a continuous Resource-leveled Longest Path of 13 tasks in sequence, with 2 additional tasks (and a split part of one of the original 13) in parallel branches.  The other 5 tasks possess from 20 to 60 days of relative float.  As a result, more of the work is both concurrent AND Critical. It turns out that the splitting of the A3 Structures task is not the key to the more condensed schedule in MSP 2016.  In fact, re-running the leveler while disallowing splits leads to a schedule that – with no splits and with a completely different sequential arrangement – finishes at the same time. The corresponding Resource-leveled Longest Path is comprised of only 8 tasks in sequence, with 1 additional, parallel/concurrent task.  This leveled schedule has fewer resource mobilizations and disruptions along the longest path while still finishing at the earliest leveled time; it appears to be lowest risk.

This example suggests that MSP 2016’s default resource leveling algorithm may be substantially more sophisticated than MSP 2010’s, and it promises – under the right circumstances – to offer shorter leveled schedules in the absence of explicit user-defined priorities.  The shorter schedules may also be accompanied by technical and resource risks associated with multiple, concurrent branches of the resource-critical-path.  Project managers using resource leveling are advised to consider and buffer these risks appropriately.

In contrast with this finding and in keeping with the prior version of this article, others have continued to observe consistently longer schedules being generated by MSP 2016 (compared to MSP 2010 and earlier), on some standardized schedule models.  These observations are noted in the comments.

What is the Longest Path in a Project Schedule?

In Project schedules, the Longest Path yields the Shortest Time.  Aside from the mental gymnastics needed to digest that phrase, the concept of Longest Path – especially as implemented in current software – has deviated enough from its origins that a different term may be needed.   

Critical Path as Longest Path

Authoritative definitions of the “Critical Path” in project schedules typically employ the words “longest path,” “longest chain,” or “longest sequence” of activities … (that determine the earliest completion date of the project.)  In other words, the path, chain, or sequence with the greatest measured length is the Critical Path.  As a rule, however, none of the associated documents are able to clearly define what constitutes the length of a logic path, nor how such length will be measured and compared in a modern project schedule.  Without a clear standard for measuring the length of something, explicitly defining the Critical Path in terms of the longest anything is just sloppy in my view.

The Original Path Length

Assessing path length used to be much easier.  In the early days of CPM (Critical Path Method) scheduling, any project schedule could be guaranteed to have ALL Finish-to-Start relationships, NO constraints, NO lags or leads, NO calendars, and only ONE Critical Path.  Under these conditions, the length of a logic path could be clearly defined (and measured) as the sum of the durations of its member activities.  Thus, the overall duration of a Project was equal to the “length” (i.e. duration) of its Critical Path, which itself was made up of the durations of its constituent activities.  That result is indicated in the figure below, where the 64-day project length is determined by the durations of the 5 (highlighted) activities on the Critical Path.  Adding up the activity durations along any other path in the schedule results in a corresponding path length that is less than 64-days – i.e. not the “longest” path. [The network diagram was taken from John W. Fondahl’s 1961 paper, “A Non-Computer Approach to the Critical Path Method for the Construction Industry,” which introduced what we now call the Precedence Diagramming Method.  Unfortunately, Microsoft Project (MSP) has an early limit on dates, so his presumed ~1961 dates could not be matched.]

Fortunately, in such simple projects, it’s never been necessary to aggregate and compare the lengths of every logic path to select the “longest path.”  The CPM backward pass calculations already identify that path by the activities with zero-Total Float/Slack, and successively “shorter” paths are identified by successively higher Total Float/Slack values.  This fact has been verified in countless student exercises involving simple project schedule networks, typically concluding with the axiom that “the Critical Path equals the longest path, which equals the path of zero-Total Float/Slack.”

Float/Slack and Path-Length Difficulties

In general, modern complex project schedules have, or can be expected to have, complicating factors that make Total Float/Slack unreliable as an indicator of the Critical Path – e.g. non-Finish-to-Start relationships, various early and late constraints, multiple calendars, and even resource leveling.  See this other article for details.  Therefore, as noted earlier, the axiomatic definition has been shortened to “the Critical Path equals the longest path.”

Unfortunately, finding the “longest path” by arithmetically summing the activity lengths (i.e. durations) along all possible logic paths and comparing the results – not easy to begin with – has gotten more difficult.  Lags, excess calendar non-working time, early constraints, and resource leveling delays all add to the true “length” of a logic path compared to the simple summation of activity durations.  On the other hand, leads (negative lags), excess calendar working-time, and the use of overlapping-activity relationships (e.g. SS/FF) reduce its length.  In addition, any hammocks, level-of-effort, and summary activities need to be excluded.  All such factors must be accounted for if the “longest path” is to be established by the implied method of measuring and comparing path lengths in the project schedule.  I don’t know of any mainstream project scheduling software that performs that kind of calculation.  Alternatively, Deep Schedule AnalysisTM using the proprietary HCP (Hidden Critical Path) Method – from HCP Project Management Consulting – appears to compute and compare the lengths of all logic paths in Primavera and MSP schedules.

Longest Path as Driving Path

Contrary to summing up and comparing logic path lengths, current notions of the “longest path” are based on an approach that does not involve path “length” at all.  As a key attribute, the longest path in a simple, un-progressed project schedule also happens to be the driving logic path from the start of the first project activity to the finish of the last project activity.  It is a “driving logic path” because each relationship in the path is “driving”, that is it prevents its successor from being scheduled any earlier than it is.  Driving relationships are typically identified during the forward-pass CPM calculations.  Subsequently, the driving path to the finish of the last activity can be identified by tracing driving logic backward from that activity, terminating the trace when no driving predecessors are found or the Data Date is reached.  The resulting driving path to project finish is also called the “longest path” even though its “length” has not been established.  This is the “Longest Path” technique that has been applied for nearly two decades by (Oracle) Primavera and adopted more recently in other project scheduling tools.

As of today, MSP continues to define Critical tasks on the basis of Total Slack, but it provides no explicit method for identifying the “Critical Path” using a “longest path” criterion.  How is the responsible MSP scheduler supposed to respond to a demand for the “critical path” when the longest path has been obscured?  Here are several options:

  1. Continue to make simple projects, avoiding all complicating factors like calendars (including resource calendars), early and late constraints, deadlines, and resource leveling. Then assume that “Total Slack = 0” correctly identifies the Critical Path.
  2. If you are using MSP version 2013 or later,
    • Ensure that your project is properly scheduled with logic open-ends only present at a single start and single finish task/milestone, then select the single finish task,
    • Try to use the “Task Path” bar highlighter to highlight the “Driving Predecessors” of your selected finish task.  In the example below, a Deadline (a non-mandatory late-finish constraint) has been applied to task Op12 in the 1961 example, and MSP has responded by applying the “Critical” flag (based on TS=0) to Op12 and its predecessors Op10 and Op2.  As a result, the Critical Path is obscured.  Applying the bar highlighter and selecting task Op18 (the project’s finish task) correctly identifies the driving path to project completion, i.e. the “longest path.”  (For clarity, I manually added the corresponding cell highlighting in the table; the bar highlighter doesn’t do that.)
    • If necessary, create and apply a corresponding filter for the highlighted bars. I’ve posted a set of macros to make and apply the filter automatically in this article.
  3. If you are using MSP version 2007 or later,
    • Ensure that your project is properly scheduled with logic open-ends only present at a single start and single finish task/milestone, then select the single finish task,
    • Try to use the Task Inspector to identify the driving predecessor of the selected task, then go to it and flag it as being part of the driving path. Repeat this until the entire driving path is marked.
    • If necessary, create and apply a filter and/or highlighting bar styles for the flagged tasks.
    • I’ve posted another set of macros to do all this (except bar highlighting) automatically in this other article.
  4. Note: The previous two approaches both rely on MSP’s StartDriver task object to identify driving relationships. As noted in this article, however, the resulting driving logic is not reliable in the presence of tasks with multiple predecessors, non-FS predecessors, or actual progress.
  5. Use BPC Logic Filter or some other appropriate add-in to identify the “longest path” in the schedule.

Whichever method or software is used, expressing the Longest Path using the Driving Path methodology has one key weakness: it has not been proved generally useful for analysis of near-critical paths.  While the Longest Path may be known, its actual length is not readily apparent.  More importantly, there is no basis for computing the lengths, and hence the relative criticality, of the 2nd, 3rd, and 4th etc. Longest Paths.  Consequently, Near-Critical paths continue to be identified based on Total Float/Slack, which is still unreliable, or – in P6 – based on unit-less “Float Paths” from multiple float path analysis.

“Longest Path” and Early Constraints

As noted several times here, the methods described for identifying the “longest path” are in fact describing the “driving path to the project finish.”  This distinction can raise confusion when an activity is delayed by an early constraint.  Consider the case below, where an activity on the longest path (Op13) has been delayed 2 days by an early start constraint.  Consequently, its sole predecessor relationship (from Op3) is no longer driving, and Op3 gains 2 days of Total Float/Slack.  As shown by MSP’s “Driving Predecessor” bar highlighter, the driving logic trace is terminated (going backwards) after reaching the constrained task.

Identical results are obtained from Primavera’s (P6) Longest Path algorithm.  This is neither surprising nor incorrect; the project’s completion is in fact driven by the external constraint on Op13, and its predecessor Op3 is quite properly excluded.

It’s clear therefore that the driving path to project completion and the longest path from the project start (or Data Date) to the project completion can differ when an early constraint is present.  P6’s “Longest Path” algorithm automatically defaults to the driving path, not the actual longest path, and to date there have been no built-in alternatives to that behavior.  As a result, some consultants suggest that P6 Longest Path analyses should be rejected when external constraints – even legitimate ones like arrival dates for Customer Furnished Equipment – are present.  (A P6 add-in, Schedule Analyzer Software, does claim to provide a true Longest Path representation in the presence of early constraints.)

BPC Logic Filter – Longest Path Filter

BPC Logic Filter is a schedule analysis add-in for MSP that my company developed for internal use.  The Longest Path Filter module is a pre-configured version of the software’s Task Logic Tracer.  The module is specifically configured to identify the project’s longest path (as driving path) through the following actions:

  1. Automatically find the last task (or tasks) in the project schedule.
    • Excluding tasks or milestones that have no logical predecessors. (E.g. completion milestones that are constrained to be scheduled at the end of the project but are not logically tied to the actual execution of the project. The resulting trace would be trivial.)
    • Excluding tasks or milestones that are specifically flagged to be ignored, e.g. (“hammocks”)
  2. Trace the driving logic backwards from the last task to the beginning of the project.
    • Driving logic is robustly identified by direct computation and examination of relative floats. (Driving relationships have zero relative float according to the successor calendar.)  The unreliable StartDriver task objects are ignored.
    • Neither completed nor in-progress tasks are excluded from the trace.
  3. Either apply a filter to show only the driving logic path, or color the bars to view the driving logic path together (in-line) with the non-driving tasks. The example below is identical to the previous one, but BPC Logic Filter formats the bar chart to ignore the impacts of the applied deadline.  The resulting in-line view is substantially identical to the bar chart of the original, unconstrained project schedule. 

BPC Logic Filter and the (True) Longest Path

As noted earlier, an early constraint can truncate the driving path to project completion.  In that case, it is debatable in my view whether the addition of non-driving, float-possessing activities into the “longest path” makes that term itself more or less useful with respect to the typical uses of the “Critical Path” in managing and controlling project performance.  Nevertheless, such an addition is easily allowed in BPC Logic Filter by checking a box.  The bar chart below shows the results of the Longest Path Filter on the early-constrained example schedule, as set up according to the driving-path (Primavera) standard.  Results are identical to those of the built-in “Driving Predecessors” highlighter in MSP (above) and of P6.

The next chart shows the complete “longest path” for the project, including the non-driving Op3 activity.

The second chart is different because the check box for “Override if successor task is delayed by constraint” has been checked in the analysis parameters form.  Checking the box causes the non-driving predecessor with the least relative float to be treated as driving, and therefore included in the Longest Path, in the event of a constraint-caused delay.

For a quick illustration, see Video – Find the Longest Path in Microsoft Project Using BPC Logic Filter.

BPC Logic Filter and Near Longest Paths

As noted earlier, the normal methods for identifying the “longest path” (i.e. the driving path) in a project schedule have not been generally adopted for analyzing near-longest paths.  P6 offers multiple float path analysis, which I wrote about here.  In addition,  Schedule Analyzer (the P6 add-in mentioned earlier) computes what it calls the “Longest Path Value” for each activity in the schedule – this is the number of days an activity is away from being on the Longest Path (i.e. the driving path to project completion.)   In the absence of demonstrated user demand, however, MSP seems unlikely to gain much beyond the Task Path bar highlighters.

BPC Logic Filter routinely computes and aggregates relative float to identify driving and near-driving logic paths in MSP project schedules.  In this context, “near-driving” is quantified in terms of path relative float, i.e. days away from driving a particular end task (or days away from being driven by a particular start task.)  Its “Longest Path” and “Near Longest Path” analyses are special cases where the automatically-selected end task is the last task in the project.  For the Near Longest Path Filter, tasks can be shown in-line (with bar coloring) or grouped and sorted based on path relative float.  The “override if successor is delayed by constraint” setting has no effect when the Near Longest Path Filter is generated.  In that case, the non-driving task will be displayed according to its actual relative float.  For example Op3 is shown below with a relative float of 2 days (its true value), not 0 days as shown on the earlier Longest Path Filter view.

Recap

  1. In the development of the Critical Path Method, the “longest path” originated as one of several defining characteristics of the “Critical Path” in simple project schedules. Specifically, the “Critical Path” included the sequence of activities with the highest aggregated duration – i.e. the “longest path”.  Actual computation and comparison of path lengths was not necessary since relative path lengths could be inferred directly from Total Float – a much easier calculation.
  2. Complicating factors in modern project schedule networks tend to confuse the interpretation of Total Float, such that it is no longer a reliable surrogate for path length. As a result, the most recent, authoritative definitions of the Critical Path tend to omit references to float while retaining references to “longest path” and, typically, logical control of the project completion date.  [Notably, the measurement and comparison of aggregated path durations (path lengths) has not been an explicit feature of any mainstream project scheduling tool, so the “longest-path” part of the definition cannot be definitively tested in general practice.]
  3. Notions of “longest-path” among current schedule practitioners are heavily influenced by the deceptively-named “Longest Path” feature in Oracle/Primavera’s P6 software. Perversely, that feature DOES NOT aggregate activity durations along any logic paths.  Rather, it identifies the driving/controlling logic path to the project’s early finish date.
  4. The “Longest Path” in P6 (i.e. the Driving Path to Project Completion) and the “longest path” (i.e. the logic path with highest aggregated duration) are NOT equivalent, particularly when the “Longest Path” is constrained by an early date constraint. There is at least one P6 add-in claiming to identify the true “longest path” (and near-“longest paths”) in this case.
  5. Microsoft Project provides several inefficient methods to identify the Driving Path to Project Completion in simple projects, but these methods are not reliable in the presence of non- Finish-to-Start relationships. There are no native MSP methods for identifying near-driving tasks nor the true “longest path” in the presence of early date constraints.  BPC Logic Filter is an MSP add-in that automatically fills these gaps.
  6. As conceived, the “longest path” criterion implied the transparent calculation and comparison of aggregated activity durations along each logic path in a project schedule. As for Total Float, however, such calculations in complex schedules have been obfuscated by complications like non- Finish-to-Start relationships, lags, and multiple calendars.  Since such obfuscation makes path lengths essentially un-testable, it appears that future Critical Path definitions should omit the “longest path” criterion in favor of a simple “driving path to project completion.”

Longest Paths in Backward Scheduled Projects (MSP) [Jan’19 Edit]

As pointed out in this recent article, the Longest Path in a backward scheduled project is essentially the “driven path from the project start,” not the “driving path to project completion.”

For more information, see the following links:

Article – Tracing Near Longest Paths with BPC Logic Filter

Video – Analyze the Near-Longest Paths in Microsoft Project using BPC Logic Filter

 

Avoid Out-of-Sequence Progress in Microsoft Project 2010-2016

[For the corresponding video presentation, see Construction CPM Conference 2020 – Avoid Out of Sequence Progress in Microsoft Project.]

Recording Actual dates that violate existing schedule logic can cause conflicts in Microsoft Project’s internal schedule calculations. The resulting Total Slack values and Critical task flags can be incorrect and misleading.  These issues are aggravated by recent (e.g. MSP 2016) versions of the software, and users are advised to minimize out-of-sequence progress.

Recording of actual progress in a logic-driven project schedule can be problematic.  As the “Actual” dates override or otherwise constrain the computed dates, the customary definitions of Float or Slack – and their resulting impacts on the “Critical” task flag in Microsoft Project (MSP) – no longer apply.  While I hope to undertake a general review of progress updating issues in a future article, this one has a special focus on out-of-sequence progress for two primary reasons:

  1. In all modern versions of Microsoft Project (e.g. ~MSP 2007+), the Total Slack values of “Critical” tasks with out-of-sequence successors can be altered unexpectedly.
  2. In MSP 2016 (and possibly beginning with MSP 2013), the Total Slack values of ALL tasks (not just “Critical” tasks) with out-of-sequence progress among their successors can be altered unexpectedly. As a result, many tasks can be shown incorrectly as “Critical” in MSP 2016 when they are not “Critical” in earlier versions.

What is Out-of-Sequence Progress

Out-of-sequence progress exists when actual progress is recorded (via, e.g. Actual Start, Actual Duration, Actual Work, %Complete, etc.) at times when the logical constraints of the schedule would normally preclude it.  For example:

  • An Actual Start is recorded for a task (the out-of-sequence, or OOS, task) whose Finish-to-Start predecessor has not finished. I.e. the Actual Start precedes the Early Start;
  • An Actual Start is recorded for an OOS task whose Start-to-Start predecessor has not started. Again, the Actual Start precedes the Early Start;
  • An Actual Finish is recorded for an OOS task whose Finish-to-Finish predecessor has not finished. Here the Actual Finish precedes the Early Finish.
  • An Actual Start is recorded for an OOS task whose Finish-to-Finish predecessor subsequently delays the completion of the work (on the OOS task).  Again the Actual Start precedes the Early Start.

In all cases there is a presumption that the recorded actual progress is more correct than the (theoretical) schedule model, so Early and Late dates are routinely overwritten by Actual dates during the schedule calculations.  When the actual progress occurs out of sequence, however, computing the Late dates (and slack values) of incomplete predecessors (during the “Backward Pass” calculations) is complicated by logical conflicts.  The software typically resolves these conflicts in a way that satisfies the needs of most users.

Completed, Out-of-Sequence Tasks

The issues discussed here are of primary concern in those cases where a task has started out of sequence and a) it remains incomplete; and b) the violated predecessor relationships remain unsatisfied (e.g. the FS predecessor remains incomplete.)  If either the OOS task or its unfinished predecessor task become complete, then they are treated like other completed tasks in Microsoft Project.  That is, they can influence the Early dates of their successors but have no impact on the Late dates of their predecessors.  As a result of this latter condition, an incomplete task whose sole successor has been completed out of sequence becomes effectively open-ended, i.e. without successors.  Under default conditions (i.e. “calculate multiple critical paths” NOT checked), the task’s Late Finish is set equal to the project’s Early Finish Date, and a high value of Total Slack is computed.

Obviously, this can have a major impact on the apparent Critical Path of a project.  In the example below (which is further described in later paragraphs), tasks CP2 and SP2 have both been completed out-of-sequence, and at a duration of only 20% of their baseline durations.  The overall project has been shortened, but the Critical Path has been truncated at CP3.  CP1 is no longer “Critical” because, in effect, it no longer has any successors.  It appears necessary to add a new FS relationship from CP1 to CP3 (and equivalently from SP1 to SP3) to re-establish the logic chain that has been broken by the completed, out-of-sequence tasks.

In Oracle Primavera P6, by contrast, the Retained Logic schedule setting automatically adjusts for activities completed out of sequence.  Thus, the need to finish CP1 before proceeding with CP3 is already included.  The result is a (presumably more realistic) 2-day delay in completion compared to the MSP result.

In-Progress, Out-of-Sequence Tasks

Key issues arise when the out-of-sequence task and its predecessor are both incomplete.  Because this behavior is sometimes different for MSP 2016 than it is for MSP 2010, we’ll look at both versions for the remaining examples.

For exploring the behavior of in-progress, out-of-sequence tasks, we examine the simple project schedule below.  The schedule is comprised of a single start milestone, a single finish milestone, a “Critical Path” string of four tasks, and a “Slack Path” string of four tasks.  The “Slack Path” is two days shorter than the “Critical Path,” with the last two tasks each having a shorter duration.  There is an unachievable Deadline applied to the finish milestone, and this creates negative Total Slack on the Critical Path.  Thus, the Critical Path tasks all have TS=-1d, and the Slack Path tasks have TS=1d.

With no progress, the project is scheduled identically in both versions of MSP.  Notably, P6 also starts with the same schedule dates.

Now let’s examine what happens when we record an out-of-sequence Actual Start on some future task.  In the example, the last task in each string (CP4 and SP4) is given an Actual Start that is one day earlier than its predecessors allow.  To keep things simple, no progress beyond the actual starts are recorded (i.e. %Complete = 0%.)  I’ve kept the “Split in-progress tasks” scheduling option checked (per default), so re-scheduling the project creates an initial split in tasks CP4 and SP4 and delays their remaining parts to satisfy the predecessor relationships.  As a result, all the tasks keep the same finish dates as before, and the project finishes on the same date as before, one day after the Deadline.

Although their start and finish dates have not changed, the logic-related information of the predecessors of the OOS tasks have been altered substantially.

  1. In both MSP 2010 and MSP 2016:
  • The Total Slack values of the Critical Path tasks that precede the OOS task (i.e. tasks CP1, CP2, and CP3) are all changed from TS=-1d to TS=0d.
  • This behavior is not justified: if the tasks are all executed according to the scheduled dates, the project will still finish one day late. The tasks should still have TS=-1d.
  • This is in fact a general result (see also Comment 1): (Presumably for MSP 2007 through MSP 365), any super-critical (TS < 0) task with an out-of-sequence, in-progress task in its successor chain will automatically have its Total Slack re-set to zero. Thus, out-of-sequence progress can cause a task with 60 days of negative Total Slack to appear much less Critical than it is.
  • This behavior can present a problem for project managers operating in a negative-slack (i.e. behind-schedule) regime, where schedule-recovery efforts are typically prioritized based on Total Slack values. Entering a single OOS Actual Start value (whether correct or not), can substantially alter the overall schedule recovery picture.
  • It seems most MSP users pay no attention to Total Slack values beyond the application of the “Critical” flag, and the observed behavior doesn’t change that. Consequently, for most users up through MSP 2010, out-of-sequence progress appears to have no substantial impact on the “Critical Path.”
  1. In MSP 2016, in addition to the prior behavior:
    • In the example, the Total Slack values of the Slack Path tasks that precede the OOS task (i.e. tasks SP1, SP2, and SP3) are all changed from TS=1d to TS=0d.
    • This behavior is also not justified. The tasks could all be delayed one day from their scheduled dates without compromising the project’s completion Deadline.  The tasks should still have TS=1d.
    • This is also a general result (see also Comment 1): (presumably for MSP 365 and maybe MSP 2013), ANY task (Critical or non-Critical) with an out-of-sequence, in-progress task in its successor chain will automatically have its Total Slack re-set to zero. Consequently, it will automatically and unavoidably be flagged as a Critical task.
    • For general MSP users after MSP 2010, therefore, out-of-sequence progress can have a substantial, even major, impact. In particular, tasks that are actually far from the Critical Path may be incorrectly flagged as Critical.
  2. The results in Oracle P6, however, appear more realistic.  The Late dates and corresponding Float values are unchanged from the initial schedule.

Below I’ve shown another perhaps more realistic example of the same simple project.  There has been a simple progress update on the Status Date of 1Oct’18, four working days into the project.  As of that date (the first Monday in the project), tasks CP1 and SP1 are still in-progress and have fallen one day behind the original plan.  Their successors (CP2 and SP2) have been allowed to start early, however, with each recording one day of actual progress.

Similar logical results are observed.  Task CP1 – the Critical predecessor of the Critical OOS task CP2 – now has TS=0d instead of TS=-1d in both software versions, and its Critical flag remains unchanged.

In MSP 2016 only, task SP1 – the non-critical predecessor of the non-critical OOS task SP2 – now also has TS=0d and is flagged as Critical.  This is incorrect.

As before, however, the late dates and Float values in the P6 version of the schedule are aligned with expectations.

Out-of-Sequence Progress and Task Path Driving Predecessors in MSP 2016

The “Task Path” bar styles provide useful methods for identifying related tasks, including the Driving Predecessors path, for any selected task in MSP 2013+.  The Driving Predecessors Task Path is particularly useful for confirming the Critical Path of a project when Total Slack is complicated by other factors.  Unfortunately, the method is not successful when out-of-sequence progress is encountered.  As shown in the figure below – repeating the two previous examples in MSP 2016 – the Driving Predecessors Task Path (orange-colored bars) is terminated when an Actual Start is reached on the backward (right-to-left) pass.  Thus, driving Task Path functionality is not compatible with out-of-sequence progress.

[The Task Path functionality is equally incompatible with in-progress schedules that involve Finish-to-Finish relationships among overlapping tasks, even if none of the progress occurs out of sequence.  Any Actual Start value terminates the progression of the associated bar formatting flag.]

Out-of-Sequence Progress and BPC Logic Filter

The impacts of out-of-sequence progress on the Total Slack of some tasks are the result of specific decisions in MSP’s backward-pass algorithms for computing Late dates and Total Slack.  Obviously, the algorithm has been tweaked between MSP 2010 and MSP 2016 leading to the even more undesired results.

BPC Logic Filter – my company’s MSP add-in for logic analysis – generally ignores MSP’s Deadlines, Late dates, Actual dates, and Total Slack values.  Instead, it performs separate backward and forward traces to determine driving logic paths and relative float values.  The latter, like Free Float, can never be negative.  Thus, when BPC Logic Filter encounters out-of-sequence progress during a trace, zero relative float is applied, and a driving relationship is inferred.  As shown in the examples below, this approach results in the correct identification of driving and near-driving paths to project completion, even when out-of-sequence progress is encountered.  [In the bar charts, the path relative float is listed to the right of each bar.  A zero-value represents the driving path with the bar characteristically colored (maroon), positive values and associated bar colors indicate the number of days away from driving the project completion.  

BPC Logic Filter also includes a Project Logic Checker to identify logic issues in Microsoft Project tasks.  Like many such tools, it automatically flags OOS tasks, along with their immediate predecessors, for correction. (Beginning with release, 1.5.5.5, the tool also differentiates the important categories of OOS tasks.  The “-I” suffix in the bar chart below identifies out-of-sequence conditions where both predecessor and successor of the violated relationship remain Incomplete; total slack in such cases cannot be trusted.)

In addition, the Logic Inspector in BPC Logic Filter automatically flags tasks with out-of-sequence progress as well as all the relationships that are violated by such progress.  Here it is shown that the out-of-sequence progress on task SP2 violates its predecessor relationship with SP1.  The late dates and total slack of the predecessor (SP1) are incorrect.

Out-of-Sequence Progress in the Real World

This examination was prompted by an associate who, after a recent “upgrade” from MSP 2007 to MSP 2016, encountered numerous unexplained changes in the “Critical Path” during project updating.  As it turned out, the updated schedule contained extensive out-of-sequence progress that explained the observed behavior.  The out-of-sequence progress was the result of a schedule model that was ultimately invalid – a poor representation of the work, either as planned or as actually executed.  As is often the case in construction, this was aggravated by the persistent executive schedule pressure that converts some technological restraints (e.g. don’t start interior finishes until the building roof and skin are closed up) from mandatory requirements into mere preferences.

A typical invalid schedule model involves the representation of multiple overlapping activities as an over-simplified Finish-to-Start string.  For example, the schedule below shows five sequentially-related, non-critical tasks, of roughly equivalent work content, that are scheduled to occur over a five-week period.  Although they are shown sequentially, it is in fact customary to execute these five tasks nearly concurrently, with each task commencing as soon as its predecessor’s progress allows – and thereafter suspending or pacing progress to match that of its predecessors.

[Some schedule purists would suggest that these tasks should be broken down into many small, repetitive work packages, all arranged with pure Finish-to-Start logic relationships.  The resulting schedule is extremely detailed and reflects a true logical plan for executing the work.  In my experience, however, such a detailed plan can often be riddled with preferential logic that is ultimately over-ruled by field decisions.  Out-of-sequence progress – in spades – is the inevitable result; the scheduling workload multiplies with no added value.

Another approach might involve five parallel tasks, each of five weeks duration, modeled with continuous relationships or at the very least with compound (joint SS + FF) relationships.  Unfortunately, neither relationship is supported by MSP.  The dummy-milestone approach that I use to mimic compound relationships in MSP seems fairly esoteric, and the common alternative – using solely SS or FF relationships – can be problematic.]

The scheduler here has chosen the most expedient route, a simplified Finish-to-Start string of five activities.  Four weekly updates of actual progress then lead to the progressed schedule shown at the bottom of the figure.  While its appearance is substantially changed compared to the original plan, the progressed schedule – in MSP 2010 – seems to correctly depict the status and slack of the multiple in-progress, out-of-sequence tasks.  Thus MSP 2010 seems largely indifferent to the consequences of invalid logic combined with out-of-sequence progress, as long as the affected tasks are non-critical.

As my associate discovered, however, MSP 2016 is far less tolerant of such practices.  When the progressed schedule is recalculated in MSP 2016, the in-progress predecessors of the in-progress, out-of-sequence tasks are shown as Critical (with TS=0).  This is incorrect, as the tasks could all slip by 15 days without delaying the project.

For comparison, here are the corresponding updates for the identical project schedule in Oracle P6 (R18.8.10).  The dates and Float values are identical to those for MSP 2010; they are correct for these non-critical tasks.  (Original and Actual Durations in P6 are handled differently than in MSP, so those are not comparable.)

Review and Recommendations

Out-of-sequence progress is an unwelcome but often unavoidable occurrence in projects with logic-driven schedules.  It happens when the actual project execution is allowed to deviate from the plan in a way that creates logical conflict in the automatic scheduling calculations.  It typically results from a combination of the following circumstances:

  1. The schedule plan includes logic relationships that are not technologically mandatory. I.e. there are a number of alternative methods available for sequencing a group of related activities, and only one (preferred) method is incorporated into the schedule.  In addition:
    • Procedural controls are inadequate to ensure that project execution conforms to the preferred logic sequence; or
    • Subsequent to initial schedule development, the preferred logic sequence is altered due to field conditions, resource limitations, or other latent factors.
  2. Subsequent to initial schedule development, technologically-mandatory logical constraints are allowed to be violated – typically under schedule performance pressure – and the resulting technical risks are accepted.
  3. The schedule is based on scope and logic definitions that are overly simplified compared to the actual or achievable plan of execution. E.g. simple Finish-to-Start relationships are used to represent overlapping, partly-concurrent activities.
  4. Subsequent to occurrence of any of the prior circumstances, the schedule logic is not revised appropriately.
  5. Schedule updates do not include regular review and correction of invalid logic in light of out-of-sequence progress.

In MSP, a task that is both started and completed out-of-sequence may cause the incomplete tasks in its predecessor and/or successor chains to become effectively open-ended, with impacts on Early and Late dates, Total Slack, and Critical task definition.  Consequently, the updated schedule may be invalid.

A task that is started out-of-sequence but remains in progress may cause substantial alterations to the Late dates, Total Slack, and Critical definitions of the incomplete tasks in its predecessor chains.  In particular:

  1. (In all modern MSP versions) Total Slack of behind-schedule tasks will change from negative to zero and in some cases turn positive. [See also comment 1.] Although the Critical flag won’t always change, any identification of driving and driven logic paths that is based on negative Total Slack will be incorrect.
  2. (In recent versions – e.g. MSP 2016) Total Slack of non-Critical tasks will change from positive to zero (with some exceptions), and each task will be incorrectly marked as Critical. [See also comment 1.]  Thus, the “Critical Path” and any other Slack-based identification of driving or driven logic paths will be incorrect.

These are pretty major consequences, yet their persistence suggests that they reflect as-designed behavior, not bugs.  It might even be suggested that the most recent “improvements” are intended to highlight the out-of-sequence progress – for correction of the associated logic.

In MSP, the only reliable way to avoid the negative consequences of out-of-sequence progress, in my opinion, is to avoid and/or minimize its occurrence.  Fundamentally, this means:

  1. Ensure that project schedules are based on sound consideration of the scope and logic of project execution.
  2. Ensure that procedural controls are put into place to a) validate, b) revise where necessary, and c) enforce the preferred sequence of activities.
  3. Where necessary, revise the schedule logic to reflect the actual/required sequence of execution.
  4. During regular progress updates, identify all out of sequence progress, and revise associated logic as appropriate to avoid the consequences noted above.

 

Relationship Free Float and Float Paths in Multi-Calendar Projects (P6 MFP Free Float Option)

This is a short article about the calculation and use of Relationship Free Float in Oracle Primavera P6, a project scheduling tool.  [It amends my previous entry on Total Float and Free Float Options in P6’s Multiple Float Path Analysis.  Both articles are further expanded (and corrected) by the technical paper presented at the 2020 Annual Meeting of AACE International.]

[The following few paragraphs are cribbed (with some edits to reflect improved understanding) from my own contribution to a discussion on Planning Planet a few years ago.]

Traditional notions of Total Float and Free Float are tied to the activities in the network, but they are not sufficient for evaluating logical float paths in complex CPM schedules, especially when variable calendars and/or late constraints are imposed.  Relationship floats are needed for identifying near-driving relationships and for multiple-float-path analyses.

Documentation seems very sketchy, but based on my own observations I believe relationship floats in P6 are calculated similarly to activity floats – that is

  1. The early and late dates of relationships are computed by treating them as activities (with single FS+0 links at each end) in the forward and backward passes through the network (Duration equals lag, normally zero).
  2. Relationship total float (RelTF) = relationship late finish (RelLF)  –  relationship early finish (RelEF); 
  3. Relationship free float (RelFF) = (Early Date of Relationship Successor Activity, ES for “FS” and “SS” links, EF for “FF” and “SF” links) – RelEF

The calendars used for the calculations seem to be as follows:

  • Early dates use predecessor calendar (from the forward pass)
  • Late dates use successor calendar (from the backward pass)
  • Relationship free float and total float use the predecessor calendar.

[Apr’19 Edit: Relationship Successor Total Float and Relationship Successor Free Float are derived from the same relationship dates, but using the successor calendar rather than the predecessor calendar.  (Figures now show both measures of relationship free float.)]

With multiple calendars, the driving and near driving paths revealed by Multiple Float Path (MFP) analysis are only partly correlated to the Relationship Free Float (and Relationship Successor Free Float) values displayed by P6.  These may require careful scrutiny to avoid misinterpretation in complex, multi-calendar projects, for the following reasons:

a) A relationship is “driving” when the successor activity possesses zero working time between the lag-adjusted predecessor and successor dates.  That is, the Relationship Free Float – according to the SUCCESSOR’s calendar (“Relationship Successor Free Float”) – is zero.

b) The Relationship Free Float that P6 displays (and that P6 appears to use as the primary path-allocation parameter for Float Paths >1, using the “Free Float” option) is computed according to the PREDECESSOR’s calendar.

Consider the simple schedule illustrated below, wherein the “Assembly” activity has three predecessors whose calendars differ from Assembly’s.  Assembly utilizes a standard 5-day calendar, while Machining 1 and Machining 2 are performed using automated equipment on a 7×24 calendar.  The Assembly Plan must be prepared during a weekly resource coordination meeting that only happens on Thursdays.

As the figure shows, the Longest Path (red bars) is comprised of the two Machining activities (finishing 12 hours apart) followed by Assembly.  As expected, both machining activities are marked as driving predecessors (there is zero relationship free float for either one according to Assembly’s calendar), and the Assembly Plan predecessor relationship is marked as non-driving.

P6’s MFP algorithm (Free Float option) allocates the three predecessors of the Assembly activity to three different Float Paths:

Float Path 1 (“most critical path”):  Machining 2, a driving predecessor, with Relationship Free Float (according to its own calendar) of 1.5 days.

Float Path 2 (“1st sub-critical path”): Assembly Plan, a non-driving predecessor, with Relationship Free Float (according to its own calendar) of 0.0 days.

Float Path 3 (“2nd sub-critical path”): Machining 1, a driving predecessor, with Relationship Free Float (according to its own calendar) of 2.0 days.

This is what the P6 Help file says for the MFP Free Float option.

Free Float – Choose this option to define critical float paths based on longest path. The most critical path will be identical to the critical path that is derived when you choose to define critical activities as Longest Path in the General tab. In a multicalendar project, the longest path is calculated by identifying the activities that have an early finish equal to the latest calculated early finish for the project and tracing all driving relationships for those activities back to the project start date. After the most critical path is identified, the module will calculate the remaining sub-critical paths.

Unfortunately, the underlined portion is not consistent with the observed behavior, where the Longest Path (i.e. the driving path to project completion) is divided between Float Paths 1 and 3.

Now we modify the schedule to give both machining activities exactly the same duration (2.5 days on a 7×24 calendar), so they both finish at 8:00 PM on Saturday.  They remain driving activities for Assembly and also have exactly the same Relationship Free Float (1.5d).  But now they trade Float Paths: Machining 1 is now on Float Path 1, while Machining 2 is on Float Path 3.

After restoring the original machining schedule (finishing 12 hours apart on Saturday), now we assign a different calendar to the one that finishes first – Machining 1 is now on a 6×24 calendar, with Sunday no longer a workday.  Consequently, the Machining 1 relationship now possesses only 1.0 days of Relationship Free Float, 0.5 days less than the Machining 2 relationship.  Nevertheless, Machining 2 stays on Float Path 1, while Machining 1 is still relegated to Float Path 3.

Several tentative conclusions seem apparent from these observations:

  1. For Float Path 1 only, the Float Path is allocated to the driving predecessor which is satisfied the latest of all the driving predecessors – without regard to Relationship Free Float.  Each of the remaining predecessors will be allocated to a new (higher-numbered) Float Path.
  2. If two or more driving predecessors finish at exactly the same time, then only one of them – selected by Activity ID, Activity Name, or some other non-logic-related criteria [Early Start, then Activity ID according to others] – will be assigned to Float Path 1.
  3. For Float Paths >1, the current Float Path is allocated to the remaining predecessor (driving or not) with the lowest Relationship Free Float of all remaining predecessors.  Each of the other remaining predecessors will be allocated to a new (higher-numbered) Float Path.  As  a consequence, legitimate members of the project’s Longest Path may be relegated to non-contiguous float paths far from the “most critical” Float Path 1.   (The fact that driving predecessors are NOT prioritized is an unfortunate weakness, in my opinion, of an otherwise robust logic analysis method.)

I’ve reviewed a number of P6 schedule submittals that seem to confirm these observations in addition to a few more:

  1. For Float Paths >1, if the remaining predecessors are driving AND have the same Relationship Free Float, then the current Float Path is allocated to the remaining predecessor that finishes latest.  Each of the other remaining predecessors will be allocated to a new (higher-numbered) Float Path.
  2. Consequently, in case of parallel driving paths, FF predecessors will be preferred (i.e. be allocated to lower-numbered Float Paths) over FS predecessors.
  3. [According to a paper first presented by Mssrs Roger Nelson and Patrick Kelly at the 2018 Annual Meeting of AACE International, and later presented at the AACE SoCal chapter meeting on 19Apr’19, the next two “tie-breakers” for path assignment are the latest Early Start date and, finally, the Activity ID.]
  4. In most cases, an ALAP-constrained predecessor automatically creates a parallel (and false) driving path with Relationship Free Float = 0.  Extensive use of ALAP constraints can lead to a proliferation of false Float Paths in the MFP results.

Don’t Confuse Critical Tasks with Critical Paths in Project Schedules

The “critical” activity flags in modern project schedules often do not correctly identify the true critical paths.  Blind acceptance of such “critical” flags to identify the critical path inhibits proper understanding, communication, and management of project schedule performance – and gives CPM a bad rap.

Basic CPM Concepts (in General):

The “critical path method” (CPM) – a ~60-year-old algorithm of fairly straightforward arithmetic – lies at the core of most modern project scheduling tools, and most project managers worthy of the name have been exposed to at least the basic CPM concepts.  Any discussion of the critical path must address the underlying conceptual basis:

  1. A CPM project schedule is comprised of all the activities necessary to complete the project’s scope of work.
  2. Activity durations are estimated, and required/planned sequential restraints between activities are identified: e.g. Predecessor task “A” must finish before successor task “B” can start, and predecessor task “C” must finish before successor task “D” can start.  The combination of activities and relationships forms a schedule logic network.  Below is a diagram of a simple schedule logic network, with activities as nodes (blocks) and relationships as arrows.
  3. Logic Relationships.  A logic relationship represents a simple (i.e. one-sided) schedule constraint that is imposed on the successor by the predecessor.  Thus, a finish-to-start (FS) relationship between activities A and B dictates only that the start of activity B may NOT occur before the finish of activity A.  (It does not REQUIRE that B start immediately after A finishes.)  Other relationship types – SS, FF, SF, which were added as part of the precedence diagramming method (PDM) extension of traditional CPM – are similarly interpreted.  E.g. A–>(SS)–>B dictates only that the start of B may not occur before the start of A.  Activities with multiple predecessor relationships must be scheduled to satisfy ALL of them.
  4. Logic Paths. A continuous route through the activities and relationships of the network – connecting an earlier activity to a later one – is called a “logic path.”  Logic paths can be displayed – together or in isolation – to show the sequential plans for executing selected portions of the project.  The simple network shown has only two logic paths between the start and finish milestones: Path 1 = (StartProject) <<A><B>> (FinishProject); and Path 2 = (StartProject) <<C><D>> (FinishProject).  [Experimenting with some shorthand logic notation: “<” = logic connection to activity’s Start; “>” = logic connection to activity’s Finish.]
  5. Schedule Calculations. Schedule dates are calculated using three essential steps:
    • During the forward pass, the earliest possible start and finish dates of each activity are computed by considering the aggregated durations of its predecessor paths, beginning from the project start milestone and working forward in time.
    • Assuming an implicit requirement to finish the project as soon as possible, the early finish of the project completion milestone is adopted as its latest allowable finish date. This can be called the finish reflection.  (Most CPM summaries ignore this step.  I include it because it is the basis for important concepts and complications to be introduced later.)
    • During the backward pass, the latest allowable start and finish dates of each activity are computed by considering the aggregated durations of its successor paths, beginning from the project completion milestone and working backward in time.
  6. Driving and Non-Driving Logic. A logic relationship may be categorized as “driving” or “non-driving” depending on its influence over the early dates of the successor activity – as calculated during the forward pass.  A driving relationship controls the early start/finish of the successor; a non-driving relationship does not.  In other words, a “driving” relationship prevents the successor activity from being scheduled any sooner than it is.  A logic path (or path segment) may be categorized as “driving” (to its terminal activity) when all of its relationships are driving.  [Such a path is sometimes called a “string.”]
  7. Total Float. In simplified terms, the difference between the early start/finish and late start/finish of each activity is termed the activity’s “total float” (or “total slack”).  A positive value denotes a finite range of time over which the activity may be allowed to slip without delaying “the project.”  A zero value (i.e. TF=0) indicates that the activity’s early dates and late dates are exactly equal, and any delay from the early dates may delay “the project.”  It is important to remember that total float/slack is nominally computed as a property of each individual activity, not of a particular logic path nor of the project schedule as a whole.  [While computed individually for each activity, the float is not possessed solely by that activity and is in fact shared among all the activities within a driving logic path.  In the absence of certain complicating factors, it is common to refer to a shared float value as a property of that path.]
  8. Critical Path. A project’s critical path is the path (i.e. the unique sequence of logically-connected activities and relationships) that determines the earliest possible completion of “the project.”  I prefer to call this the “driving path to project completion.”  Other logic paths through the schedule are considered “near-critical paths” if they are at risk of becoming the critical path – possibly extending the project – at some time during project execution.  In our simple project shown below, the critical path is Path 1, whose total duration of 4 weeks (20 days on a standard 5dx8h calendar) controls the early finish of the completion milestone.

    In unconstrained schedule models incorporating only a single calendar (and without other complicating factors), the finish reflection causes the activities on the critical path to have late dates equal to their early dates; i.e. TF = 0.  Consequently, any delay of a critical-path activity cascades directly to delay of the project completion.  The near-critical paths are then defined as those paths whose activities have TF more than zero but less than some threshold.  In traditional “critical path management,” activities that are NOT on or near the critical path may be allowed to slip, while management attention and resources are devoted to protecting those activities that are on or near the critical path.  More importantly, acceleration of the project completion (or recovery from a prior delay) may only be accomplished by first addressing the activities and relationships on the critical path.

[Note: The definition of “critical path” has evolved with the introduction of new concepts and scheduling methods over the years.  The earliest definitions – based on robust schedule networks containing only finish-to-start relationships, with no constraints, no lags, and no calendars – were characterized by the following common elements:

  • It contained those activities that determined the overall duration of the project (i.e. the “driving path to project completion.”)
  • It contained those activities that, if allowed to slip, would extend the duration of the project (hence the word “critical”).
  • A delay of any of its activities would be directly transmitted to an equal (matching) delay of the project completion.
  • Its activities comprised the “longest path” through the schedule network. That is, the arithmetic sum of their durations was greater than the corresponding sum for any other path in the network.
  • After completion of the forward and backward passes, its activities could be readily identified by a shared total float value of zero.  Thus TF=0 became the primary criterion for identifying the critical path.

With the incorporation of non-FS relationships, early and late constraints, lags, and calendars in modern project scheduling software, these observations are no longer consistent with each other nor sometimes with a single logic path.  Some of these inconsistencies are addressed later in this article.  Only the first of these defining elements (“driving path to project completion”) has been generally retained in recent scheduling standards and guidance publications, though implied equivalence of the others continues to persist among some professionals.]

Software – the Critical Activities / Critical Tasks:

The basic element of modern project schedules is the activity or task.  In most scheduling tools, logic paths are not explicitly defined.  Nevertheless, the obvious importance of the critical path dictates that software packages attempt to identify it – indirectly– by marking activities that meet certain criteria with the “critical” flag.  Activities with the “critical” flag are called “critical activities” (or “critical tasks”) and are typically highlighted red in network and bar-chart graphics.

Applying Critical Flags using Default Total Float Criteria

The simplest criterion for flagging a task as “critical” is TF=0.  This is the primary method that most new schedulers seem familiar with, and it is the default criterion for some software packages.  As noted earlier, this criterion is applicable to schedules with no constraints and only a single calendar.  In Microsoft Project (MSP) and Oracle Primavera P6 (P6), the default “critical” flag criterion is TF<=0, and the threshold value of “0” can be adjusted.  The differences between these criteria and the simpler TF=0 criterion are justified by four primary concerns:

  1. Risk Management. Due to the inherent uncertainty of activity duration estimates, the critical path of a real-world project schedule – as ultimately executed – often includes an unpredictable mix of activities from the as-scheduled critical path and near-critical paths.  In the absence of quantitative schedule risk assessment, it is reasonable to consider all such (potentially-critical-path) activities equally when evaluating project schedule risks.  This purpose is easily served by applying the “critical” flag to all activities whose TF value is less than or equal to some near-critical threshold.
  2. Late Constraints. Overall project completion priorities (and contractual requirements) often lead to the imposition of deadlines (in MSP), late-finish constraints (in MSP and P6), or project constraints (in P6).  Such constraints can override the finish reflection and cause the late dates of some activities to be earlier or later than they would be in the absence of the constraints.  As a result, total float can vary among the activities on the driving path to project completion.   In a project with multiple constrained milestones, the driving path to only one of them (the most “urgent”) can be expected to have a constant total float value (i.e. the lowest total float.)  Due to intersecting logic paths, total float can vary along the driving paths to other constrained milestones.   Applying the “critical” flag to activities with total float less than or equal to the project’s lowest total float marks those activities that are on the driving path to the most urgent constrained milestone in the project.  If a project constraint (in P6 only) is applied, the lowest total float value may be greater than zero; without a more urgent constraint, the marked activities then denote the driving path to the final activity in the project.
  3. Negative Float. Late constraints can cause late dates to precede early dates for certain activities.  This results in negative values for total float/slack (i.e. TF<0).    In practically all cases, negative total float indicates that the activity cannot be scheduled in time to satisfy one or more of the deadlines or constraints (though which one of these is violated may not be clear); and some corrective action is necessary.  [*The concept of negative float – and the constraints that create it – were not included in the foundations of CPM and PDM.  Negative float is not universally accepted among scheduling professionals today, and not all scheduling software supports its calculation.]

    Applying the “critical” flag to all activities with total float less than or equal to zero then marks all activities that:

      • Are on the driving path to an unconstrained project completion (i.e. TF=0, controlled by the project’s finish reflection); or
      • Are on the driving path to a constrained project completion or intermediate milestone that is just barely met (i.e. TF=0, controlled by deadline/constraint); or
      • Are on the driving path to project completion where an explicit project completion milestone is violated (i.e. TF<0, controlled by project deadline/constraint); OR
      • Are on the driving path to some intermediate activity whose constraint is violated (i.e. TF<0, controlled by intermediate deadline/constraint); or
      • Are on any number of non/near-driving paths to one or more constrained project completion or intermediate milestones, (i.e. TF<0). Though non-driving, these paths must still be shortened (in addition to shortening the driving and nearer-driving paths) to meet the milestones.

     

  4. Working-Time Calendar Effects. When activities with different calendars are logically connected in a schedule network, the interval between the finish of a predecessor task and the start of its successor may sometimes contain working time for the predecessor but not for the successor.  If this occurs, then a driving logic relationship exists, but the predecessor still has room to slip without delaying any other tasks or the project (i.e. it possesses float.)  More generally,  when a driving logic path contains activities with different calendars, the interval between the early and late dates of one activity may contain more or less worktime than the corresponding intervals of its mates.   Thus, total float may vary along a single driving logic path, including the critical path.  The amount of this variation depends on the size of potential offsets between calendars: from a few hours (for shift calendar offsets) to a few days (for 5-day and 7-day weekly calendars offsets) to a few months (for seasonal-shutdown calendar offsets).

    Applying the “critical” flag to all activities with total float less than or equal to the largest calendar-related offset will mark all activities that:

    • Are on the driving path to project completion with TF<=0;
    • Are on the driving path to project completion but with TF>0 (and less than the specified offset);
    • Are NOT on the driving path to project completion but have TF less than the specified offset. These are false positives.  For these activities, total float could be controlled either by the finish reflection (TF>=0) or by some other constraint.

Critical Flags and Critical Paths

Unfortunately, applying the “critical” flag as noted for most of these considerations has one consistent result:  the continuous sequence of activities and relationships constituting a “critical path” often remains obscured.  It is disappointing that the majority of project schedulers – using MSP or P6 – continue to issue filtered lists of “critical” activities as “the critical path.”  Much of the time – especially in MSP – they are not.  Even among expert schedulers, there is a persistent habit of declaring total float as the sole attribute that defines the critical path rather than as a conditional indicator of an activity’s presence on that path.

When an activity is automatically marked “critical” based on total float/slack, the primary conclusion to be drawn is simply, “this activity has total float/slack that is at or below the threshold value.  That is, there is insufficient working time available between the early- and late- start/finish dates.”  If total float/slack is less than zero, then one might also conclude, “this activity is scheduled too late to meet one or more of the project’s deadlines/constraints.”  [If automatic resource leveling has been applied, then even these simple conclusions are probably incorrect.]  These are important facts, but a useful management response still requires knowledge of the driving logic path(s) to the specific activities/milestones whose deadlines/constraints are violated – knowledge that total float/slack and its associated “critical” flag do not always provide.

Workarounds for Total Float Criteria

P6 provides several features, not available out-of-the-box in MSP, for correctly identifying the critical path when total float criteria do not.  Specifically:

  1. For Risk Management. P6’s multiple-float path analysis (MFP) allows the identification of successive driving and near-driving paths to specified project completion milestones.  Monitoring progress on these paths is worthwhile for risk management.  I’ve previously written about MFP analysis HERE.  P6 does not support using float paths (the output of MFP analysis) as an explicit criterion for the “critical” activity flag.
  2. For Late Constraints and Negative Float. P6 allows a negative critical float threshold.  It is possible to set this threshold low enough so that only the “path of lowest total float” is marked as critical.  In the absence of working time calendar effects, this criterion can be effective in identifying the (most) critical path.  Thus it is possible to correctly identify the project’s critical path when: a) there is only a single constraint on the project; and b) that constraint coincides with the sole project completion milestone; and c) that constraint is violated (creating negative float).
    • MSP does not allow a negative critical float threshold, so correct identification of the critical path in a negative float scenario is not possible. All tasks with negative total slack are automatically and unavoidably flagged as “critical.”
    • If the P6 schedule has a project “must finish by” constraint, then the activities on the critical path may have positive total float. In that case, the lowest-float criterion may be applied (using a positive threshold) to correctly identify the critical path.
  3. For Working-Time Calendar Effects. Unlike other project scheduling software, P6 allows the “critical” activity flag to be assigned on the basis of some criterion other than total float – called Longest Path.  The name is misleading, as the method is based on driving logic rather than activity durations.  Any activity that is found on the driving logic path to project completion is flagged as “critical.”  (The algorithm tracks driving logic backward from the task(s) with the latest early finish in the project.)  The Longest Path criterion ignores the total float impacts of multiple calendars and constraints.  While it is effective in identifying the project’s critical (logic) path, Longest Path alone is not useful for identifying near-critical paths.  MFP analysis (noted above) is useful for this purpose.  “Longest path value ™,” a relative-float metric available in Schedule Analyzer Software (a P6 add-in) also helps to identify near-critical paths in these circumstances.  For a more detailed review, see What is the Longest Path in a Project Schedule?

MSP provides no out-of-the-box solutions to address these weaknesses in critical path identification.  Total float/slack remains the sole basis for applying the “critical” flag, yet the impacts of constraints, deadlines, and calendars remain unaddressed.  In MSP 2013 and later versions, the “task path” bar style modifier does provide a basis for graphically identifying the driving path to a selected completion activity, and this is helpful.  Nevertheless, a logic tracing add-in (like the BPC Logic Filter program that I helped to develop) is necessary to correctly identify the controlling schedule logic – including the true critical path – in a complex MSP schedule.

Definitions and Recommended Practices

Defense Contract Management Agency (DCMA – 2009)

DCMA’s in-house training course, Integrated Master Plan/Integrated Master Schedule Basic Analysis (Rev 21Nov09) is the source of the “14-Point Assessment” that – because its explicit “trigger” values are easily converted to pass/fail thresholds and red/yellow/green dashboards – is seen as a de-facto industry standard for schedule health assessment.  The course materials contain the following definitions:

(Slide 28) Critical Path ~ Sequence of discrete work packages that has the longest total duration through an end point.
~ has the least amount of total float
~ cannot be delayed without delaying the completion date of the contract (assuming zero float).
(Slide 98) Critical Path – Definition: a sequence of discrete tasks/activities in the network that has the longest total duration through the contract with the least amount of float.
~ A contract’s critical path is made up of those tasks in which a delay of one day on any task along the critical path will cause the project end date to be delayed one day (assuming zero float).
(Slide 99) The critical path is ‘broken’ whenever there is not a sequence of connected critical path tasks that goes from the first task of the schedule until the last task.  A broken Critical Path is indicative of a defective schedule. 

These definitions are mostly (though not entirely) consistent with each other.  They do share a common emphasis on the … “longest”… “sequence” … with “lowest total float” and direct transmission of delay from any critical-path task directly to the project’s completion.  Obviously, the reliance on total float makes them incompatible with any project schedule that incorporates multiple calendars, late constraints, or resource leveling.

(Slide 97) Critical Task:  Some tasks possess no float…they are known as critical tasks.
~Any delay to a critical task on the critical path will cause a delay to the project’s end date.

Unlike most of the later definitions, DCMA’s appears to contemplate the existence of critical tasks that are not on the critical path.  Obviously, the expectation that such critical tasks possess “no float” is not compatible with negative-float regimes, nor is it compatible with the positive-float regimes that accompany project “must finish by” constraints in P6.

AACE International (2010 & 2018)

AACE International (formerly the Association for the Advancement of Cost Engineering) maintains and regularly updates its Recommended Practice No. 10S-90: Cost Engineering Terminology.  The most recent issue of RP 10S-90 (June 2018) includes the following definitions:

CRITICAL PATH – The longest continuous chain of activities (may be more than one path) which establishes the minimum overall project duration. A slippage or delay in completion of any activity by one time period will extend final completion  correspondingly. The critical path by definition has no “float.” See also: LONGEST PATH (LP). (June 2007)

CRITICAL ACTIVITY – An activity on the project’s critical path. A delay to a critical activity causes a corresponding delay in the completion of the project. Although some activities are “critical,” in the dictionary sense, without being on the critical path, this meaning is seldom used in the project context. (June 2007)

Unfortunately, these definitions fall apart in the presence of multiple calendars, multiple late constraints, or negative total float – when the second and third clauses in both definitions no longer agree with the first.  They appear distinctly out of sync with modern project scheduling practices, and (according to AACE International’s Planning and Scheduling Subcommittee Chair) an update is pending.

AACE International’s RP No. 49R-06, Identifying the Critical Path (last revised in March 2010) instead defines the Critical Path as

…the longest logical path through the CPM network and consists of those activities that determine the shortest time for project completion.  Activities within this [group (sic)] or list form a series (or sequence) of logically connected activities that is called the critical path. 

Aside from the apparently inadvertent omission of a word, I don’t have any problem with this definition.  It is certainly better, in my opinion, than the first.

RP 49R-06 notes the existence of “several accepted methods for determining the critical path” and goes on to describe the four “most frequently used” methods:

  1. Lowest Total Float. This is as I described under Workarounds for Total Float Criteria, above.  Although this method is listed first, the RP spends four pages detailing the issues that make total float unreliable as a CP indicator.  As long as the CP is to be defined only with respect to the most urgent constraint in the schedule (including the finish reflection) – and there are no calendar issues –  then this method provides a useful result.
  2. Negative Total Float.  In apparent acquiescence to the limitations of MSP, the RP describes this method by first abandoning the fundamental definition of the critical path as a specific logic path.  It then allows the “critical” classification for any activity that must be accelerated in order to meet an applied deadline or constraint.  Ultimately, the RP attempts to justify this method based solely on certain legal/contractual considerations of concurrent delay.  It is not useful for those whose primary interest is timely completion of the project, or a particular part of the project, using critical path management principles.
  3. Longest Path.  This “driving path to project completion” algorithm, as I described above in Workarounds for Total Float Criteria, has been implemented in versions of (Oracle) Primavera software since P3 (2.0b).  It is the preferred method for P6 schedules with constraints and/or multiple activity calendars.  A similar algorithm is included in BPC Logic Filter, our add-in for Microsoft Project.  While the method is nominally aimed at finding the driving path(s) to the last activity(ies) in the schedule, it can be combined with other techniques (namely a super-long trailing dummy activity) to derive the driving path to any specific activity, e.g. a specific “substantial completion” or “sectional-completion” milestone.
  4. “Longest Path Value.”  This is an expanded method for identifying the driving and near-driving paths to project completion.  The method works by adding up relationship floats leading to a specific substantial completion milestone.  If the aggregate value of these floats along a specific logic path (i.e. “Longest Path Value”) is zero, then that path is identified as the critical path.  While the RP suggests that this method can be performed manually (presumably by “click-tracing” through the network of a P6 schedule), manual implementation in complex schedules is tedious and error prone.  As implemented in Schedule Analyzer Software, this method is essentially an improved version of  P6’s Longest Path method (except that the add-in cannot change the “critical” flag for activities.)  It is a preferred method in P6 for those possessing the Schedule Analyzer Software.  BPC Logic Filter performs similar analyses – using “path relative float” instead of “Longest Path Value” – for MSP schedules.

While not listed among the “most frequently used” methods, P6’s MFP analysis option is briefly addressed by the RP in the context of identifying near-critical paths.  BPC Logic Filter performs similar analyses for MSP schedules.

None of the four methods described are useful for identifying the resource critical path (or resource-constrained critical path) of a leveled schedule.

Project Management Institute (PMI-2011)

PMI’s Practice Standard for Scheduling (Second Edition, 2011) explicitly defines the critical path as…

Generally, but not always, the sequence of schedule activities determining the duration of the project.  Generally, it is the longest path through the project.  However, a critical path can end, as an example, on a schedule milestone that is in the middle of the schedule model and that has a finish-no-later-than imposed date schedule constraint.

Unlike the RP (49R-06) from AACE International, PMI’s Practice Standard provides no meaningful method for quantitatively identifying the activities of the critical path (or any logic paths) in a particular schedule model.  In fact, in its description of the precedence diagram method (PDM – the modern version of CPM used by most modern scheduling software) the Practice Standard acknowledges the complicating factors of constraints and multiple calendars but notes that “today’s computerized scheduling applications complete the additional calculations without problems.”  Then it concludes, “In most projects the critical path is no longer a zero float path, as it was in early CPM.”  The Practice Standard goes on to scrupulously avoid any explicit link between total float and the critical path.  The impact of all this is to just take the software’s word for what’s “critical” and what isn’t.  That’s not particularly helpful.

Finally, educating senior stakeholders on the subtle difference between “schedule critical” and “critical” is always one of the first issues faced when implementing systematic project management in non-project focused organizations.  The Practice Standard’s several conflicting definitions of critical activities tend to confuse rather than clarify this distinction.

U.S. Government Accountability Office (GAO-2015)

The GAO’s Schedule Assessment Guide: Best Practices for Project Schedules (GAO-16-89G, 2015) has been taken to supersede the earlier DCMA internal guidance in many formal uses.  (Nevertheless, the GAO’s decision to discard any formal trigger/threshold values – a good decision in my view – means that the DCMA-based assessments and dashboards remain popular.)  The GAO document contains the following formal definitions:

Critical path: The longest continuous sequence of activities in a schedule. Defines the program’s earliest completion date or minimum duration.

[With some minor reservations related to meaning of “longest,” I believe this is a good definition.]

Critical activity: An activity on the critical path. When the network is free of date constraints, critical activities have zero float, and therefore any delay in the critical activity causes the same day-for-day amount of delay in the program forecast finish date.

[Unfortunately, the caveats after the first clause are insufficient, ignoring the complicating effects of multiple calendars.]

For the most part – and despite the float-independent formal definition above – the Schedule Assessment Guide’s “Best Practices” tend to perpetuate continued reliance on total float as the sole indicator of the critical path.  In fact, “Best Practice 6: Confirming That the Critical Path Is Valid” does a good job of illustrating the complicating factors of late constraints and multiple calendars, but this review leads essentially to the differentiation of “critical path” (based on total float alone) from “longest path” (based on driving logic).  This is a direct contradiction of the formal definition above.  In general, the text appears to be written by a committee comprised of P6 users (with robust driving/Longest Path analysis tools) and MSP users (without such tools.)  Thus, for every “longest path is preferred,” there seems to be an equal and opposite, “the threshold for total float criticality may have to be raised.”  This is silly.

National Defense Industrial Association (NDIA-2016)

The NDIA’s Integrated Program Management Division has maintained a Planning & Scheduling Excellence Guide (PASEG), with Version 3.0 published in 2016.  The PASEG 3.0 includes the following key definitions:

Critical Path: The longest sequence of tasks from Timenow until the program end. If a task on the critical path slips, the forecasted program end date should slip.

Driving Path(s): The longest sequence of tasks from Timenow to an interim program milestone.  If a task on a Driving Path slips, the forecasted interim program milestone date should slip.

The second clause of each definition – which presumes a single calendar – is included in the Schedule Analysis chapter but is excluded from the formal definition in Appendix A.  Timenow is effectively the data/status date.  The PASEG does not define or mention critical task/activity as distinct from a “task on the critical path.”

The PASEG notes, “Some of the major schedule software tools have the ability to identify and display critical and driving paths. Additionally, there are many options available for add-in/bolt-on tools that work with the schedule software to assist in this analysis.”  [I suppose BPC Logic Filter would be one of the mentioned add-in tools for Microsoft Project.]

The PASEG also mentions some manual methods for identifying critical and driving paths, e.g.:

a. Imposing a temporary, super-aggressive late constraint and grouping/sorting the output (presumably by total float and early start.  Though not explicitly mentioned in the method description, total float is the key output affected by the imposed constraint.)  Obviously, this method isn’t reliable when more than one calendar is used.

b. Building a custom filter by manually “click-tracing” through driving logic and marking the activities.  This method is most reliable in P6, with some caveats.  It is reliable in MSP only under some fairly restrictive conditions.

In general, these methods are non-prescriptive, though the emphasis on driving logic paths (rather than total float) seems clear.

Guild of Project Controls (GPC, “The Guild” – 2018)

The Guild is a relatively young (~2013) international community of project controls practitioners – initially associated with the PlanningPlanet.com web site – whose founding members have assembled a Project Controls Compendium and Reference (GPCCaR).  The GPCCar takes the form (more or less) of an introductory training course on Project Controls, including Planning and Scheduling.  The GPCCaR includes no formal Glossary, Terminology, or Definitions section, so “critical path” and “critical path activities” accumulate several slightly varying definitions in the applicable Modules (07-01, 07-7, and 07-8).  In general, “zero total float” and “critical path” are used interchangeably, and the complications of multiple calendars and multiple constraints in P6 and MSP are ignored.  This is not a suitable reference for complex projects that are scheduled using these tools.

American Society of Civil Engineers (ASCE)

ASCE Standard ANSI/ASCE/CI 67-17 – Schedule Delay Analysis is one of the few documents with a clear and correct distinction between the critical path and the collection of critical activities:

Critical path—The series of logically connected tasks that define the minimum overall duration for completion of the project, also known as the longest path. There can be more than one critical path in the schedule.

Critical activities—Activities with zero or negative float in a schedule reflecting a current adjusted completion date, some of which may not be on the critical path.

Recap

  1. A full understanding of driving and non-driving schedule logic paths for major schedule activities is useful for managing and communicating a project execution plan.
  2. The most important logic path in the project schedule is the “critical path,” i.e. the driving path to project completion.  Overall acceleration (or recovery) of a project is only made possible by first shortening the critical path.  Acceleration of activities that are not on the critical path yields no corresponding project benefit to project completion.  Multiple critical paths may exist.
  3. Some traditional notions of critical path path attributes – e.g. critical path activities possess no float; slippage or acceleration of critical path activities always translates directly to project completion – are not reliable in modern project schedules.
  4. Total float remains a valuable indicator of an activity’s scheduling flexibility with respect to completion constraints of the project.  An activity with TF=0 may not be allowed to slip if all project completion constraints are to be met.  Activities with TF<0 must be accelerated if all the constraints are to be met.
  5. Project scheduling software typically defines individual activities as “critical” without fully accounting for common complicating factors like multiple constraints and calendars.  As a result, the collection of “critical” tasks/activities in a complex project schedule often fails to identify a true critical path.
  6. A critical task/activity is best defined (in my opinion) as either:
    1. An activity that resides on the critical path; or
    2. An activity whose delay will lead to unacceptable delay of the project completion; or
    3. An activity whose delay will lead to unacceptable delay of some other constrained activity or milestone.
    4. In general, these conditions are mutually exclusive, and different activities within a single project schedule may satisfy one or more of them.
  7. Professional project managers and schedulers should be careful not to automatically characterize “critical” tasks (i.e. those with low total float) as indicators of a project’s critical path when complicating factors are present.

 

The Right Way to Construct Level of Effort (LOE) Activities in Primavera P6

This short entry demonstrates why Level of Effort activities in Oracle Primavera P6 should be constructed using predecessor successor relationships only – especially when designating Critical activities using the Longest Path algorithm. 

P6 LOE Activities

Level-of-Effort (LOE) activities in Oracle Primavera P6 project scheduling software are useful for summarizing the schedule dates of other (primary) activities.  They are effectively a replacement for the “hammock” activities that existed in prior versions of Primavera products, including Primavera Project Planner (P3) and SureTrak Project Manager.

In prior software versions, hammocks were constructed by establishing a group of start-controlling activities as SS predecessors and a separate group of finish-controlling activities as FF successors.  To avoid circular logic, the same activity could not be included in both predecessor (i.e. Start) and successor (i.e. Finish) groups.

In P6, LOE activities can be constructed using practically any mix of predecessors and successors, and the help file implies that the two can be used almost interchangeably.  Although the same activity may not be included as both a predecessor and a successor, it MAY be included multiple times (e.g. as both an SS and FF predecessor) in either group.  This allows for more flexibility in modeling activities that are not strictly hammocks.

P6 Help – Level of effort activity

Here’s a portion of the LOE Help topic in P6, which seems indifferent to the types of relationships used in specifying primary activities.

A level of effort activity is similar to but different from a hammock activity.

  • A level of effort activity uses its assigned calendar to summarize its dates. Hammocks are not scheduled using their own calendar.
  • Any type of relationship can be assigned to a level of effort activity. Only a start-to-start and finish-to-finish relationship can be assigned to a hammock activity.
  • A level of effort activity’s duration is calculated from the earliest early start of its predecessors/successors (linked to the start end of the level of effort activity) to the latest early finish of its predecessors/successors (linked to the finish end of the level of effort activity).A hammock activity’s duration is calculated from the earliest early start of its predecessors to the latest early finish of its successor activities.

Problems with LOEs and Longest Path

Despite the increased flexibility afforded by LOE activities in P6 – and its implied indifference to predecessor or successor relationships – many users continue to use SS-predecessors and FF-successors when establishing LOE dates.  This can create issues when the project’s Critical Path is defined using the Longest Path option.

Here is a simple example project comprising a Critical Path of activities A-B-C-D-E-F-G with a side branch of associated activities A1-B1-C1-D1-E1-F1.  The side activities must be completed before the final activity G can start.  All activities are on the same calendar, and there are no constraints or resources.  The corresponding Critical Path, as determined by Total Float (TF=0), is highlighted red on the Gantt Chart.

Critical Path by Total Float

For this simple project, the Critical Path based on Total Float should be identical to the Critical Path based on “Longest Path” – i.e. the “driving path to project completion.”  Unfortunately, this is not the case, as activities A1 (TF=10) and B1 (TF=8) are now included – incorrectly – on the Longest Path.

Critical Path by Longest Path (Driving Logic)

This error is caused by the Level of Effort activity LOE-1, which summarizes the dates of the side activities in Branch 1.  Similar to P3 hammocks, this LOE activity is constructed using SS-predecessors to govern its start and FF-successors to govern its finish.  The resulting dates are correct, but two complications are introduced.

P3 Hammock-Type Relationships
  1. As a rule, P6 marks all relationships to and from LOE activities as “Driving,” even when the relationship does not control any of the LOE’s dates.
  2. The Longest Path algorithm, which traces driving logic backward from the project completion, does not differentiate LOE driving relationships from other driving relationships during the trace.

As a result of these complications, the Longest Path calculation traces driving logic backward from activity F1 through LOE-1 to its two driving predecessors A1 and B1.  The latter two – along with their entire chains of driving predecessor logic, if any – are now included on the Longest Path.  P6 automatically removes the LOE activity and its relationships from the Longest Path (after the backward pass), but the “Critical” flag on its “driving” predecessors remains.

Activities that are incorrectly included on the Longest Path may be identified by first noting those “Critical” activities whose successor relationships are NOT BOTH “Driving” and “Critical.”  Then their “driving” predecessors are traced until either 1) there are no more driving predecessors, or 2) an activity is reached that has a separate successor relationship that is BOTH “Driving” and “Critical,” provided that this successor does not ultimately lead to another incorrect LOE predecessor.  Such an examination can be tedious.

A quicker identification is obtained by re-calculating the schedule using Multiple Float Paths (Free Float option), with no End activity specified.  Unlike the erroneous LP algorithm, MFP analysis correctly truncates the backward trace at LOE activities.  Consequently, the true driving path to project completion (i.e. the correct Longest Path) of our simple project is identified as “Float Path 1,” and activities A1 and B1 are correctly relegated to Float Paths 6 and 5, respectively.

Multiple Float Path (FF)

When analyzing Multiple Float Paths under the Longest Path regime: if a “Critical” activity has a Float Path that is higher than the Float Path of even one non-critical activity (all computed using the Free Float option), then that “Critical” flag may be incorrect.*

[Edit:  Unfortunately, similar symptoms are produced by MFP analysis when P6 incorrectly assigns legitimate members of the Longest Path to much higher Float Path numbers.  See details in Relationship Free Float and Float Paths in Multi-Calendar Projects (P6 MFP Free Float Option).]

Avoiding the Problem

The simplest way to avoid this Longest Path LOE bug is to avoid including a mix of predecessors and successors when specifying the primary references for LOE activities.  That is, use ONLY predecessors or ONLY successors.  Since LOEs are technically inheriting their dates entirely from the primary activities, my initial preference was to use only predecessor relationships in the LOE.

Using Predecessor-Only Relationships

As shown here, the Longest Path of our simple example project is fixed by re-constructing LOE-1 logic using predecessors only.  The two activities that were formerly “FF” successors of the LOE are now “FF” predecessors of the same LOE.  The dates and float calculations are essentially unchanged, but the Longest Path is now correct.

Predecessor-Only Relationships – LOE-1

Sometimes we want to make an LOE activity whose finish corresponds to the finish date of the project.  Creating such an LOE using predecessor-only relationships doesn’t work as intended, however, because P6 picks the new LOE as the starting point of its Longest Path trace during the backward pass.  Thus, our intention to stop logic flow through the LOE is foiled.  In the next figure, this is shown by the new LOE-2 activity, whose finish is determined by the finish milestone (FM) of the project.  Since LOE-2 is the latest-finishing activity with no successors, P6 starts the backward pass with it, ultimately including its non-critical predecessors A1 and B1 (incorrectly) on the Longest Path.

Predecessor-Only Relationships – LOE-2

Using Successor-Only Relationships

In contrast, if LOE-2 is constructed using successor-only relationships, then P6 will never choose it as the starting point of the Longest Path trace.  Moreover, the absence of predecessors will ensure that when P6 does encounter such an LOE during the backward pass, the Longest Path trace will not continue past the LOE.  As shown below, constructing LOE-2 using successor-only relationships leads to dates, float calculations, and Longest Path calculation that are all correct.  This example suggests that successor-only relationships should be the preferred method for specifying LOE activities in P6.*

Successor-Only Relationships – LOE-2

*Thanks to astute reader A Lou Gonzalez, who pointed out the special issue with using predecessor-only relationships to define LOE activities at the project completion date.

Late Cost Curves for LOE Activities

While using predecessors-only to specify LOE activity dates seems to fix the logic-flow issues, it has no effect on another defect of LOE activities.  As Wail Menesi, has described in his LinkedIn Pulse entry,  P6 incorrectly computes the remaining late start dates of certain in-progress LOE activities.  As a consequence, the “Remaining Late” resource distribution for the affected activity is skewed to the left of the data date.  In other words, P6 indicates that certain incomplete work must be completed in the past to avoid delaying the project, even when there is no negative float.  That’s clearly incorrect.  Fortunately, the issue seems to be limited to LOE activities who’s start dates are determined by Finish-to-Start relationships – a relatively rare structure in practice.