How to Estimate
p2 schedules best when you provide realistic low and high estimates in hours. Use ranges to capture uncertainty, not just a single number.
In p2
- Low Estimate - Optimistic estimate in hours
- High Estimate - Pessimistic estimate in hours
Estimate With Ranges
- Low is a realistic best case, not a miracle.
- High is a realistic worst case, not a disaster scenario.
- The person doing the work should own the estimate.
- Ranges should be wide when the work is unclear and narrower as you learn.
Target An 80% Confidence Interval
- Treat your low and high as an 80% confidence interval.
- You should be wrong about 1 out of 5 tasks. That is normal and expected.
Calibration check: Imagine a wager where you win if the task finishes between your low and high, and lose if it finishes outside the range. Compare that to a simple 80% bet (like roulette with 8 winning numbers out of 10). If you would always take your range bet, the range is too wide. If you would always take roulette, the range is too narrow. This calibration exercise is adapted from Douglas W. Hubbard’s How to Measure Anything.
Break Work Down
- Break work into tasks small enough to estimate well, typically 2-8 hours.
- If you cannot estimate a task inside that range, split it further.
- When you need tighter estimates, break work down further, often into 1-6 hour tasks.
- Smaller tasks are easier to estimate. Painless Software Schedules is a good primer.
- Estimates improve with practice. How to Measure Anything (Douglas W. Hubbard) discusses calibration practice.
Start Wide, Then Refine
- When you lack context, start with a wide range and narrow it as you explore.
- If the schedule needs more certainty, invest time to reduce uncertainty and tighten the range.
- Clear requirements help: sketches, mockups, or quick prototypes often reduce estimation error.
Keep Estimates Current
- Update estimates as requirements or priorities change.
- As you learn, adjust the remaining estimate to keep schedules accurate.