The Fairytale of Alex in Blunderland
Alex is a traditionally-trained Project Scheduler who performs traditional Planning and Scheduling practices dutifully, and flawlessly. Yet, the fruits of his labor are – well, flawed! How is this possible? It is possible because the underlying theory behind standard Planning and Scheduling practices is itself misguided in many important and cumulatively-impacting ways. In this blog, we will consider how some of these factors invalidate our interpretation of the typical Schedule Update. Specifically, we will consider three commonly-overlooked observations.
Observation #1: Durations are Not as Precise as We Thought
Let me start by asking a simple-enough question: how long is one day? While Project Schedules can be expressed in other units of time (e.g., hours, weeks, months, or years), the overwhelming majority of them, especially in construction, use a “day” as the smallest unit of time. So, that is why I ask you: how long is a day?
- Clock Hours: According to the clock it is twenty-hours long. But when an Activity has a duration of “one day,” we do not expect the workers to slave away for 24 hours, do we? So, the clock definition cannot be the correct answer.
- Work Shifts: Another, more likely, explanation is that we use the word “day” to refer to a work shift – maybe one or more work shifts. We might imagine that the default assumption is that “one day means one shift.” But since this is a default we might also define a “workday” as comprised of more than one shift.
But then, we have to consider that different members of the Project Team might work different numbers or lengths of shifts. One contractor might work a ten-hour shift, while others work eight-hour shifts. Some contractors might work back-to-back shifts for the duration of their presence on the project, or perhaps just for a short while in order to get back on schedule, if they had fallen behind. And even if all of the subcontractors agreed to conform to the General Contractor’s workday and work shift standards, the General Contractor might, from time to time, call for extended workdays.
We Schedulers Traffic in Workdays.
Let’s face it: Everything about the Critical Path Method schedule reduces to measures of workdays.
- Duration: The Activity duration aggregates intended workdays.
- Calendars: Automated Work Calendars distinguish workdays from non-workdays.
- Dates: All of the calculated Earliest and Latest Dates in the CPM schedule are derived by adding and subtracting workdays.
- Float: Total Float and Free Float are expressed in workdays.
- Paths: Whether critical, near-critical, or non-critical, criticality is measured by Total Float, which is expressed in workdays.
What all of this means is that Project Execution is a blurry mass of conflicting and juxtaposed workdays. And since the Project Schedule is a only a virtual/digital simulation of the intended Project Execution Strategy, then the Project Schedule is a corresponding swirl of time units that, as it turns out, are inconsistently defined.
Recalling the Activity Duration Profile is Not a Flat Line
In an earlier blog (“Realistic Schedules Require Prudent Expectations“), I noted that,
“The general assumption is that an Activity is performed in a steady level of intensity, commencing at the very first moment of the Activity and lasting until the very last moment of the Activity [and] in turn, the next (follow-on) Activity is assumed to start a millisecond after its predecessor has finished. This is not how things work in the real world.”
I went on to explain that:
“for any given trade and for any given installation process there is typically a ramp-up period at the start of the Activity, as well as a ramp-down period at the end of the Activity. In fact, Momentum Science (part of Cognitive Project Management) identifies three distinct stages in every Activity.
- Production Stage: This is the primary portion of the Activity when the Work Scope is being performed. It is called a “production” stage because effort during this stage produces the Work Scope of the Activity.
- Pre-Production Ancillary Stage: This is a typically short “ramp-up” period of time needed to achieve positional and conditional readiness prior to performance during the Production Stage.
- Post-Production Ancillary Stage: This is a short “ramp-down” period of time required to demobilize the Activity’s workforce as well as to clear and reset the workspace as a courtesy to the next trade in line to perform work in the same area.
This is hardly a minor matter. What happened to the Ancillary Stages of an Activity’s performance? When does our Schedule contemplate it taking place? Is it assumed to happen during the length of the Activity? Or, as I strongly suspect, is it assumed to be “taken care of” between the Activities? [As an aside, I happen to think that this issue — whether, as a matter of Schedule Development Policy, the Ancillary Stages of Activities are to be included within the Activity Duration, or not —is one of the most important oversights in our current understanding of the Project Scheduling science. I encourage the Project Management community to conduct further research into this “black hole” within the Project Time Management universe.
Activity Durations are Rounded
Next, let us not forget that as a practical matter Activity Durations are always subject to rounding. More than any other reason one can think of, the fact that Activity Durations are rounded should temper our silly notion that Project Scheduling is some precise science with temporal tolerances “down to the day.” Yet there we are, presenting (or defending) to Owners how small changes in Total Float of as little as one or two days can be interpreted to mean something profound about how the project will turn out, X weeks, months, or years later!
Rounding is a big deal because not all work divides evenly by eight (setting aside the additional matter of breaks, lunch, and as just discussed Ancillary Stages). All legitimate books on Critical Path Method scheduling advocate the same general formula for deriving Activity Duration values. They are calculated by factoring the rate at which work is expected to be performed divided into the amount of work to be performed. I can assure you that this formula only rarely yields a whole number with no remainder. No, raw Activity Duration calculations result in values like 12.2, 5.3, 7.1, 2.3, 9.8, and so forth. So, if we follow the recommended derivation for Activity Durations, what are we supposed to do with 5.2 hours, 11.7 hours, or 34.45 hours? Do we round down, or round up? Either way, haven’t we just lost some of our precision?
Estimates of Current Activity Status
While we are talking about rounding, let us consider another pocket of imprecision that affects our Scheduling Practices. Think about the recommended best practices involved in updating the Project Schedule. In particular, think about how we “status” a single Activity that has been started, but not yet completed. For this we have are two available measures of Activity Progress.
Percent Complete: This procedure looks at the portion of the Activity that has been performed, and estimates how much work was accomplished.
Remaining Duration: This procedure looks at the portion of the Activity that has not been performed, and estimates the amount of time still needed to complete the Activity.
Aren’t both of these values just estimates? And, continuing our observation about rounding, don’t we round these values, too? Have you ever encountered a Percent Complete estimate expressed in tenths of a point? And aren’t all Remaining Durations rounded, just as their Original Duration cousins were, back when the schedule was first being developed?
Ancillary Stages as Percent of Overall Activity Duration
I’d like you to imagine a situation where we have two Activities. Suppose a Pre-Production Stage of one-day and a Post-Production Stage of 0.50 days for these two Activities. Now imagine that Activity A has a four-day Activity Duration, while Activity B has a 12-day Activity Duration. If we assume that the Ancillary Stages are assumed to be contained within the Activity Durations, then can you picture in your mind how their respective Activity Duration Profiles might look? The Activity Profile for Activity A would show a leveled Production Stage of 2.5 days. As a profile, it would look short and tall, like squared-off bell curve. Yet, Activity B would be long and short, like a short-legged Dachshund, with a leveled Production Stage of 10.5 days.
Why the Activity Duration Profile matters is because we know (thanks to ICS-Research) that the majority of schedule delay occurs during the Ancillary Stages. It follows, then, that short-duration Activities have a higher probability of slippage than their long-duration Activity counterparts.
And yet, I would be remiss if I did not also note that this conclusion (the higher probability of slippage) can be often offset by the rounding effect discussed earlier. Taken together, perhaps these two points explain why the law of averages helps to level out the net differences between what was planned and what actually occurs, in terms of individual Activity temporal performance.
If nothing else, these insights should lend support to the intuitive practice among prudent contractors to slightly pad their Activity Durations. Frankly, I am not too bothered by a modest amount of Activity padding, especially on smaller-duration Activities. But again, we must be careful not to get too carried away with the amount of padding used. It takes very little padding, compounded across scores of Activities, to entirely destroy the integrity of an otherwise excellent Project Schedule.
Observation #2: Schedules Represent Commitments, Not Predictions
As has been discussed in other blogs, the Project Schedule as a model of Project Execution Strategy is not a prediction of how things will turn out. Instead, it is a set of commitments by and between those who are responsible for performing the project’s work.
Project Schedule: Set of Commitments
There should be no dispute that a Project Schedule is a set of commitments. Yet, believe it or not, there is anything but consensus on this point among seasoned planners and schedulers. My view is that each Activity in a schedule represents as many handshakes as there are arrows (logic ties) connecting to that Activity! These relationships – just as the Activity Durations and resultant Activity Dates – are not merely goals, they are promises. One man’s word to another.
Activity Durations: Temporal Placeholders
Viewed in this light, Activity Durations should be seen as temporal placeholders. Each subcontractor honors his word and arrives at the time and place of his Activity. There, he sees what number of “days” he had previously agreed to during Schedule Development Logic Sessions. Armed with this information, and in keeping with his work ethic, come hell or high water he struggles to complete his Activity in the number of days allowed (that he set for himself).
[Note: It should be noted that any non-working hours between Activities should be reserved for the exclusive use of subcontractors performing work prior to non-work periods (Owner and contract permitting). This way, the follow-on trades can count on (a) their day starting on time, and (b) their Activity starting on the right day. This non-work “make-up” period is the spacing between the dominos written about in a previous blog. “The Art of Stacking Dominoes.”]
Observation #3: We Routinely Ignore the Human Factor
The third observation about how the Conventional Wisdom misunderstands the composition of the Project Schedule — and what can be gleaned from a routine updating of it — is the ridiculousness of using probability theory to speculate on the degree by which a Project Schedule might be overrun. To the point: Any statistical method that perceives Activity performance as somewhat akin to rolls of the dice or flips of a coin completely disregards the human factor: will and determination.
Probabilities work quite well on problems that do not involve the human mind. For instance, imagine dropping a shovel full of river gravel into a sieve box that has five layers of different-sized openings. After only a few shovels-full, we could begin to form some fairly credible predictions as to what percent of the stones would be stopped at each level. To further make my point, let me present another statistical problem. Except, this time, we will consider it from two opposite perspectives: one without the employment of human judgment, the other with.
- The Situation: Picture a mile-long road which, along one side, has a continuous parking lane made up of 100 parking spaces of varying sizes (between 16-24 feet in length, each). Now imagine a long caravan of 100 vehicles of varying sizes (also between 16-24 feet long, each) that approaches this one-mile stretch of road.
- Scenario 1: Assume that the 100 vehicles proceed alongside the parking spaces, such that the first vehicle in line stops adjacent to the farthest-most parking space, and each other vehicle comes to stop alongside a parking space, as well. If the car could fit, the driver is to park the car in the space adjacent to it. Question: What percent of the 100 parking spaces would be filled with a car?
- Scenario 2: In this scenario, we make the same assumptions, except — we allow the drivers to pick the space that would seem adequate for their vehicles size. We would only need to impose two simple rules. (1) Never pick a space larger than you need. (2) Start with the largest vehicle first and proceed until the smallest vehicle chooses last. Now, applying these two rules, Question: What percent of the 100 parking spaces would be filled with a car?
The point I am trying to make is that an inherent flaw in so many statistically-based approaches to scheduling (Monte Carlo, risk analysis, etc.) is that they completely dismiss the human factor. They treat Activities like steel balls in a pinball machine, or coins being slipped or dice being tossed. [Note: Beyond the scope of this blog, I will parenthetically note that these statistical approaches also typically ignore the uniqueness of Activities. That is, no two Activities are the same, even if they have the same Activity Duration. Remember our earlier discussion about Activity Duration Profiles that helps us understand this.]
I hope this blog is sufficient to give you pause the next time you are faced with interpreting a Schedule Update. Please keep in mind the inherent imprecision of the CPM science, its great power and potential notwithstanding. To Scheduler, Project Manager, and Owner alike — let’s not go ballistic if Total Float changes by a few days from one Schedule Update to the next.