Home > Topics

A Method For Estimating Work

Method

For each task, record a P5 and a P95 estimate for how much time you’d expect to pass before the work is done if it were yours to complete.

Definitions

1. Task
2. P5 estimate - a soft lower bound. In the long run, a calibrated estimator completes 5% of estimated tasks before their P5 estimate for that task.
3. P95 estimate - a soft upper bound. In the long run, a calibrated estimator completes 95% of estimated tasks before their P95 estimate for that task.
4. Time
5. Done - task specific, well defined, testable criteria indicating that a task is complete

Execution

Calibration

When a calibrated estimator has completed 20 estimated tasks

1. exactly 1 task was completed before their P5 estimate for that task
2. exactly 1 task was completed after their P95 estimate for that task
3. the duration of the remaining 18 tasks fell between their corresponding P5 and P95 estimates
   Of course, this is also true in percentage terms, regardless of the population of tasks considered.
   Depending on the nature of the work, context, and the estimator’s experience this may be easy or hard to achieve.
   An estimator should observe their task durations relative to their estimates and regularly review their overall performance.
   Observing how our completed task durations differ from our estimates informs how we should adjust our judgements.

Miscalibrations

1. Optimistic upper bounds. More than 5% of tasks are completed after our P95 estimates.
2. Optimistic lower bounds. Fewer than 5% of tasks are completed before our P5 estimates.
3. Pessimistic lower bounds. More than 5% of tasks are completed before our P5 estimates.
4. Pessimistic upper bounds. Fewer than 5% of tasks are completed after our P95 estimates.
5. Unjustified precision. Too many durations fall outside of our estimates on either side.
6. Extreme case: P5 = P95
7. Our estimates are never correct.
8. Unjustified uncertainty. Too many durations fall inside our estimates.
   1. Extreme case: P5 = 0, P95 = infinity
   2. Our estimates are always correct.

Improving Estimates

1. Break down complicated tasks
2. Imagine a world where this task is done.
3. Imagine the world when this task is started.
4. Imagine doing the work required to complete this task.
5. Write down the steps.
6. Write down P5 and P95 estimates for each step.
7. Review your task performance relative to your estimates regularly.
8. When completing a task, reflect on surprises.
   1. What took more or less time than you expected?

Teamwork

1. A second estimator opens the door to learning and collaboration. When estimates differ significantly
   1. Maybe there’s something one estimator isn’t considering
   2. Maybe there’s something one can learn from the other
2. Breaking down work
   1. is often much faster than completing the task
   2. invites collaborative planning. Improving a plan is often faster, easier, cheaper, and feasible compared to improving a completed task.
   3. produces a reference plan that someone else may use for a similar task in the future.
   4. highlights processes that may be learned from or improved upon