How to forecast a realistic delivery date from your board (instead of guessing)

Humans estimate the work itself and forget everything around it: the waiting, the handoffs, the half-day a card sits in review before anyone looks. So a job that's "two days of work" takes a week of calendar time, and the estimate wasn't wrong about the effort, it was wrong about the flow. The cure isn't trying harder to guess; it's measuring how long similar things actually took, start to finish, last time.

Cycle time is the calendar time a card takes from when work starts to when it's done, waiting included. Because Production Board timestamps every stage move, you can read the real cycle times of past cards instead of inventing a number. Look at the last 10 to 20 similar jobs and you'll see a spread, say most finished in 5 to 9 days. That spread, not a single hopeful figure, is your forecast. The data already disagrees with your gut, and the data is right.

A new job doesn't start the moment it's requested; it starts when the team gets to it. So a realistic delivery date is cycle time PLUS the wait for everything already ahead of it in the pipeline. If five cards sit in front of this one and the team finishes roughly two a week, that's a two-and-a-half-week wait before work even begins. Forecasts blow up when people quote the cycle time and silently ignore the queue. Count the work in progress ahead of the new card and add its drain time — finding your bottleneck tells you which stage is likely holding the queue longest.

Never give a single date; give a range with a confidence, "likely 2 to 3 weeks, and I'll know more once it clears the queue". A range is honest about the spread real work has, and it protects you from the false precision that makes a missed single-date estimate feel like a broken promise. Re-check the forecast as the card moves: once it's actually in progress, the remaining cycle time tightens the range fast. A forecast is a living read of the board, not a one-time guess.