Critical path method challenge: Activity duration estimation

Critical path method challenge: Activity duration estimation

The critical path method includes several standard steps. While there are different recommendations in which order these steps need to be performed, it is a general acceptance that one step is important: the duration of each activity has to be accurately estimated.

The activity estimation is a process itself. The process could be very simple.

“Using past experience or the knowledge of an experienced team member, you must now estimate the time required to complete each activity.”

or more complex:

“The duration is an estimate of how long it will take to accomplish the work involved in the activity. In many cases, the number of resources that are expected to be available to accomplish an activity, together with the productivity of those resources, may determine the activity’s duration. A change to driving resource allocated to the activity will have an effect on the duration, but this is not a simple “straight-line” relationship. Other factors influencing the duration are the type or skill level of the resources available to undertake the work, resource calendars, and the intrinsic nature of work. Some activities will take a set amount of time to complete regardless of the resource allocation.”
Practical Standard for Scheduling, PMI

How can an estimator/s accurately provide estimates for the duration of ONE activity? Is it actually possible?

Firstly, they need to understand the work represented by the activity in the schedule. Ideally, the Definition of Done (DoD) for each activity has to be developed, but it is rarely the case. Usually, estimators rely on activity and WBS names only.

Example: “Complete Design” could mean:

  • 1st draft available
  • Design reviewed
  • Design updated
  • Design submitted for review board
  • Design endorsed
  • Design fully approved

Then, estimators need to understand the level of uniqueness of this activity. It may vary from ‘exactly the same work was done before many times’ to ‘this work is completely new to us’.

Then, estimators need to think about matters that may impact the duration:

  • What is the Volume of work to be achieved?
  • Which skills are needed?
  • Which resources have these skills?
  • Could additional skilled resources be hired?
  • What is the minimum (technically) required number of resources?
  • What is the maximum (technically) possible number of assigned resources?
  • What productivity (volume units per hour) does each resource with the required skill have?
  • If more than one skill is required, which skill has the primary impact on the duration?
  • When could the activity be (technically) performed?
  • How may external work impact the availability of resources?
  • How may other project activities impact the availability of resources?
  • What type of materials are required?
  • How may materials supply impact the duration?
  • How may project’s environmental factors impact the duration?
  • How may external factors impact the duration?

Translating these questions to a scheduling language: there are project parameters that may impact activity duration. Some of these parameters could be well-known or they may have avoidable (epistemic) or unavoidable (aleatory) uncertainties (U). Also, some of these parameters are interrelated (I) with other activities.

  • Volume of work (U)
  • Skills (resource pool)
  • Renewable (Labour and Equipment) Resource Quantity (U, I)
  • Consumable (Materials) resources quantity (U, I)
  • Resource Productivity (U)
  • Activity Calendar (U)
  • Resources Calendars (U)
  • Resource availability (A)
  • Technical constraints (U)
  • Threats (I)
  • Opportunities (I)
  • Shift calendars (U)
  • Shift quantity (U)

It is usually easy to understand how each parameter impacts the activity duration but it is not easy to understand the cumulative effect. The human brain is simply not able to handle so many parameters simultaneously. If George A. Miller’s statement is correct#, our brains can handle only 7 (+/-2) objects simultaneously.

# The Magical Number Seven, Plus or Minus Two: Some Limits on our Capacity for Processing Information

It is important to mention that not all activities have all mentioned above parameters, or a value could be obvious and even ignored:

  • Activities may not include any materials
  • Only one resource with standard working hours is required
  • An activity technically could be performed at any time.

There are different project techniques that could help with estimations:

  • Top-Down
  • Bottom-Up
  • Expert judgment
  • Comparative (or analogous)
  • Parametric model
  • Three-point

These techniques help to get accurate estimations but they have to be applied to primary parameters that impact durations, not directly to durations.


Our project received a new bulldozer. It differs from other bulldozers used in organisations, so we don’t have any statistical data to understand the productivity rate. Project decided to use comparative analysis (or Expert judgment) as a planned rate and calculate activity durations based on this rate. When they receive actuals, they can update the productivity rate and recalculate all future activities with the assigned of this bulldozer. Also, if the productivity rate is stored in Corporate Reference Book, they could update other projects where this bulldozer is planned or use the findings for future assignments.

If activity durations were collected instead, it would be very difficult to recalculate and update all related activities, as they would need to be done individually. Also, historical data is (almost) useless. While we know that actuals were different from planned durations, there is no data in the schedule that tells us “why”, and it is hard to reuse historical data for future projects.

Expert judgment

Expert judgment often is only available for project teams as an estimation technique.

Unfortunately, expert judgment is not a reliable method. Daniel Kahneman in his book “Thinking Fast And Slow” explained that when people need to find a solution to very complex problems, they are likely to use a “fast thinking system” that could give them quick but not necessarily accurate answers. What is even worse, the answers are inaccurate and inconsistent. If the same expert provides activity estimations in a week, it is likely that his estimations would not be the same. Different cognitive biases may impact their opinion. In some cases, provided estimations are going to be very conservative. In other situations, they will be too optimistic.

What is the alternative

Stop asking estimators to provide most likely activity durations and ask them to provide parameters that impact this duration instead.

If a parameter has an uncertainty, usually it is much easier to provide a range or 3 points (optimistic, probable, pessimistic) rather than a single number.

Then all available data could be used to calculate the optimistic, probable, pessimistic and planned durations!

This approach has some advantages:

  • Easier identification of why an activity wasn’t performed as planned
  • Collected actuals could be used for planning for future activities even if the activities are not exactly the same:

    * Activity requires the same skill but has a different volume of work.

    * Activity has the same volume of work and skills but will be performed by different resources.

    • Some of the activity parameters are applicable to many activities, so once collected the data could be used for many activities
    • While a project may have thousands of activities, it doesn’t require thousands of skills. Even a project-driven portfolio has a small and stable list of required skills.
    • Easier identification of project and portfolio skills bottleneck
    • Each resource has a very limited number of skills. It could be beneficial to help critical resources gain a new appropriate skill or improve already existing skill
    • 3 points estimations could be used for more sophisticated probabilistic methods: Monte Carlo Simulation, ‘3 scenarios’, etc.
    • Result of the Monte Carlo Simulation depends on the simulation parameter. The result of the simulation is more accurate when primary uncertainties are used instead of the secondary parameters (like duration and cost)
    • Structured low-level data could be used for AI
    • It addresses some cognitive biases. Just a few to mention:

    * Optimism bias

    * Representativeness bias

    * Confirmation bias

    * Framing effect

    * Parkinson Law

    * Availability heuristic

    Alex Lyaschenko

    PMO | Portfolio Planning & Delivery | PMP | P3O Practitioner | AgilePM Practitioner | Six Sigma