All jobs / Project / delivery manager
Estimation interview questions
Estimation questions a Project / delivery manager can be asked, from first job to senior. Open one to see the strongest answer and why it works.
A stakeholder asks “when will it ship?” You have a rough slice and unknowns. Strongest answer?
The strongest answer: A range with what would move the ends — “sooner if the dependency lands, later if it does not” — not a single Tuesday I cannot defend
Why it works: Exactly. A range plus the movers. A single date you cannot defend is a promise, not a forecast.
- Give a range, not a Tuesday.
- Name what moves the ends.
- Do not pick the midpoint as the date.
- Do not invent certainty.
When would you T-shirt size work instead of estimating it in points or days?
The strongest answer: Early and coarse — a roadmap of many items to compare and sequence, not commit; S, M, L, XL is fast and honest about how little we know, and items are refined later
Why it works: Right — the coarseness is the feature: it matches the uncertainty at that stage and stops false precision.
- Coarse early, refine later.
- Use sizes to compare and sequence.
- Refine before commitment.
- Never size to dodge a range.
A director asks for “our velocity number” so they can forecast the year. What do you not do?
The strongest answer: I do not invent a points-per-sprint figure I cannot show from a stable, honest history — I offer a range from what we have actually finished, or I say the history is not there yet
Why it works: Exactly. No invented velocity. Observed finishes, as a range, or an honest “not yet.”
- Do not invent a velocity.
- Use observed finishes, as a range.
- Say when history is too thin.
- Forecasting is allowed; fiction is not.
A new teammate asks what a “5” means in hours. Strongest answer?
The strongest answer: It is relative to our other slices, not a timesheet — I will not convert the board into hours I cannot defend
Why it works: Exactly. Relative size. Hours-as-points are a second lie.
- Points compare slices.
- Do not publish an hours chart.
- Do not bend the number to the sprint.
- Estimating is allowed; fiction is not.
You have a few months of finished slices and a new, lumpier ask. How do you forecast?
The strongest answer: I show a range from what we have actually finished and I widen it for the new lumpiness — I do not quote a single week as if the past were a law
Why it works: Exactly. Observed finishes, as a range, widened for new uncertainty. No fake precision.
- Start from observed finishes.
- Keep a range; widen for new work.
- Do not multiply tickets into a date.
- Do not wait for a perfect breakdown.
A stakeholder needs a time estimate next week for a piece of work nobody on the team has done before.
How do you produce it?
The strongest answer: I break it down with the people who will do it, estimate each part optimistic, likely and pessimistic, give the total as a range with its assumptions, and spike the biggest unknown
Why it works: Exactly — decomposition, three-point estimates and a spike are how you turn “no idea” into an honest range.
- Decompose, three-point, range.
- Estimate with the builders.
- State the assumptions.
- Spike the biggest unknown.
Stakeholders accept your range but want it narrower by next week. What do you offer?
The strongest answer: A specific check that would narrow it — a spike, a dependency answer, a slice in a usable build — not a promise to pick a Tuesday by Friday
Why it works: Exactly. Narrowing needs new information. Offer the check, not a costume date.
- Offer a check that narrows.
- Do not shrink the range with no news.
- Do not donate a Tuesday.
- Do narrow when you learn.
How do you forecast a lump of work without inventing a velocity?
The strongest answer: From the distribution of what this team has actually finished, as a range, and I say when the work is unlike that history — I do not invent a points-per-sprint law
Why it works: Exactly. Observed finishes. A range. A caveat when the work is unlike the past. No invented law.
- Use this team’s finished history.
- Keep a range.
- Say when the work is unlike the past.
- Do not invent a velocity law.
A team wants to stop estimating entirely and forecast by counting finished items.
When does that work, and what do you insist on?
The strongest answer: When items are right-sized — a few days at most — and throughput is fairly stable; I insist on a size ceiling, splitting anything over it, and forecasting in ranges from the count
Why it works: Right — counting is a good forecast only when the things counted are similar in size; the ceiling is what makes them similar.
- Count only right-sized items.
- Set a size ceiling.
- Split what exceeds it.
- Forecast in ranges from throughput.
A key person leaves and a dependency slips in the same week. What do you do to the forecast?
The strongest answer: I widen the range now, I name the two shocks, and I say when I will re-narrow — I do not keep the old range to look stable
Why it works: Exactly. Shocks widen ranges. Stability-as-silence is a lie.
- Widen the range now.
- Name the shocks.
- Say when you will re-narrow.
- Do not keep the old range for optics.
Finance wants a single date for the annual plan. History is thin and the work is new. Strongest stand?
The strongest answer: I give a wide range, I say what would narrow it, and I will not sign a single date I cannot defend — I do not invent a Tuesday for the model
Why it works: Exactly. Wide range, narrowing conditions, no costume date. Models do not get to create knowledge.
- Wide range when history is thin.
- Name what would narrow it.
- Do not sign an indefensible date.
- Do not keep a private second forecast.
Dates keep slipping and people have learned to give a single number they do not believe. What culture do you set?
The strongest answer: That a defendable range is the professional answer, that narrowing has a check, and that a costume date is a defect — I do not reward the bravest Tuesday
Why it works: Exactly. Ranges are the culture. Rewarding bravery is how you buy fiction.
- Reward defendable ranges.
- Narrowing needs a check.
- Treat costume dates as defects.
- Do not reward the bravest Tuesday.
A PMO wants a company velocity so finance can model the year. What do you refuse?
The strongest answer: I refuse a single invented velocity for unlike teams — I offer per-team ranges from observed finishes, or I say the histories do not add
Why it works: Exactly. No company velocity. Unlike histories do not add. Per-team ranges or an honest “cannot add.”
- No company-wide invented velocity.
- Per-team ranges from finishes.
- Say when histories do not add.
- Do not fit the annual model.
You can set one forecasting rule for the company. What is it?
The strongest answer: Public dates are ranges with named movers and a time we will narrow them — a single Tuesday without that is not an official forecast
Why it works: Exactly. Make the range the policy. A Tuesday without movers is unofficial.
- Ranges with movers are the policy.
- Name when you will narrow.
- A bare Tuesday is not official.
- Do not force a one-week width.