Estimating Timelines
Rohan asks Meera when UPI autopay can ship. The honest answer is "I don't know yet", but the tempting answer is a date that sounds confident. Whatever she says will end up on a slide for the leadership team, so it had better mean something.
The rule
Give a range with its assumptions, never a single hopeful date.
An estimate is a forecast, not a promise. A single date hides how uncertain it is; a range tells the reader how far to rely on it.
- Break the work into small pieces. Pieces of a few days each are far easier to estimate than one big blob, and breaking the work down drags hidden work into the light.
- Let the people doing the work estimate. Meera asks questions such as "what could make this take longer?", but she doesn't supply the numbers.
- Count the unglamorous work. Code review, testing, security review, rollout and the bugs you will find on the way all take time.
- Give a range and name what it assumes. "Four to six weeks, assuming the bank's test environment is ready by Monday and two engineers stay on it full time."
- Check against history. If the last payment-gateway integration took seven weeks, a three-week estimate for the next one needs a very good explanation.
- Update the forecast. As pieces finish, the range narrows. When it moves, tell people straight away.
People reliably underestimate, even when they know their past projects overran. Small pieces, ranges and history are the correction.
The shortcut
Break it, range it, name the assumptions. Three moves, and the answer comes out as a sentence someone can actually use.
Break it
Arjun and Neha split the work: bank API integration 4–6 days, mandate screens 3–5, webhooks and retries 5–8, testing in the bank's sandbox 3–6, staged rollout 3–5.
Notice what the shortcut prevents. Quoting only the low end, 18 days, would sound great in the meeting and become a missed date a month later. The range carries the uncertainty honestly, and the assumptions tell Rohan exactly what to watch.
Drills
Answer fast — trust the rule.
Which estimate is most useful to Rohan?
Three pieces, done one after another, are estimated at 2–3 days, 4–6 days and 1–2 days. What is the total range?
Who should produce the first estimate for a piece of work?
Two weeks in, the bank's sandbox is late and the range slips by a week. What should Meera do?
Remember
- Give a range with its assumptions, never a single hopeful date.
- Break it, range it, name the assumptions.
- Update the forecast, and say so the moment an assumption breaks.