Estimating as a team: a 20-minute routine
Estimates don't have to be precise to be useful. They have to be consistent. Here's a short routine teams use to get there.
Most estimates go wrong in the same way: one person guesses alone, under pressure, and the number sticks. The fix isn't a better formula. It's a short, regular habit where the team estimates together.
The routine (20 minutes, once a week)
- Pick the next ten tasks that don't have an estimate. Not the whole backlog.
- Read each one aloud in one sentence. If nobody can say what "done" means, it goes back to its owner.
- Everyone estimates silently in hours: 2, 4, 8, or 16. No other numbers.
- Reveal together. If the numbers are close, take the higher one. If they're far apart, the highest and lowest explain their reasoning in one minute each, then vote again.
- Split anything over 16. A task that big hides at least two tasks.
Why a fixed scale. Choosing between 4 and 8 is easy. Choosing between 5 and 6 is a waste of time. The scale also makes bars on the Gantt comparable at a glance.
Why take the higher number. Teams are consistently optimistic. Taking the higher of two close estimates corrects for that without anyone having to argue for padding.
Check the routine monthly. Compare estimates with actual time on finished tasks. If the team is consistently off by the same ratio, that's fine — forecasts can use the ratio. If it's random, the tasks are probably too big.
The routine works in any tool. In TaskD, estimates set bar lengths on the Gantt, so a new estimate immediately shows whether the plan still fits.