The estimate becomes a commitment the moment it leaves the building
An estimate that goes into a proposal stops being an estimate. It is lifted into a commercial schedule, sometimes verbatim, and it becomes the basis on which a delivery team is measured against a margin they were not consulted about. That is why the internal sign-off question is a real interview question rather than a process detail. What an interviewer is testing is whether you understand that your output is a liability transferred to colleagues, and whether you have the nerve to insist somebody accepts it before it goes.
The failure has a recognisable shape. Presales builds a number under deadline pressure, circulates it for comment, receives silence from delivery because everyone is busy on live work, treats silence as agreement, and submits. Nine months later the engagement is over-running and the review concludes that the estimate was optimistic — a finding about a document, which is easier than the finding about the process that let it out unowned.
Sign-off means a named person, not a distribution list
The concrete answer is that three things have to be true before submission, and each has an owner.
Somebody who will actually run the work has to accept the effort. Not a practice head signing off a spreadsheet, but the person who will be accountable for delivering it, or the nearest available proxy if staffing is not decided. Somebody commercially responsible has to accept the price, which is a different decision from the effort and should be visibly separated from it — margin, rate and risk appetite are theirs, effort is not. And somebody has to accept the assumptions, because the assumptions are where the risk actually lives and they are the part that gets edited out for readability.
That last point deserves emphasis in the room. An estimate is a function of assumptions, and reviewing the number without reviewing the assumptions is theatre. When a reviewer says the number feels high, the productive question is which assumption they would change, and by how much.
When the delivery lead says no
Refusal is not one behaviour, and treating it as one is where junior answers go wrong. Diagnose it first.
She may be right. Her objection may rest on an integration you have priced from a datasheet, a data migration you have sized from a table count, or a customer-side dependency you have assumed will be met. This is the most common case and the answer is simply to change the estimate.
She may be protecting herself against a plan she does not trust, which is a different objection wearing the same clothes. Her real concern is that she will be held to a fixed number under a shape of contract that leaves her no room, and she is expressing it as a challenge to the arithmetic. The fix is not more effort, it is a different commercial structure or a bounded first phase.
Or she may be refusing to engage at all, because she is delivering three other things and the deal is not her problem this week. This is not a technical disagreement and cannot be resolved by discussing the estimate. It is a resourcing question for her manager and it should be raised as one, quickly, because it does not improve on its own.
| Reason for the refusal | What resolves it | What does not |
|---|
| A specific item is under-sized | Re-estimate that item with her assumption stated | Adding a percentage across the total |
| She distrusts the contract shape | Phase the commitment, or change the pricing model | Reassuring her it will be fine |
| Scope is genuinely unknowable | A paid discovery phase, or a priced range with a named trigger | Choosing a midpoint to move things along |
| No capacity to review | Her manager, today | Interpreting silence as approval |
Do not average a disagreement
The tempting resolution when presales says one number and delivery says a substantially larger one is to submit something between them. It has the appearance of pragmatism and it produces the worst of both positions: a number too low to deliver and too high to win, with nobody who believes it.
The disciplined alternative is to find the specific line items where the two views diverge, because it is almost never spread evenly. Usually two or three items carry nearly all of the gap, and they diverge because of a factual disagreement — how many interfaces, whether the data is clean, whether their environment exists — which can be tested rather than debated. If the disagreement survives that examination, it belongs in the proposal as a stated assumption with a consequence attached, which is honest and also protective. If it cannot be tested before submission, it is a risk somebody more senior than either of you has to accept knowingly.
Escalation is a document, not a confrontation
If the gap does not close, the answer is to escalate, and how you do it separates senior candidates from vocal ones. Put it in writing, briefly, to whoever owns the commercial risk. State the two numbers, the items that differ, why each side holds its view, what you would need to resolve it and what you recommend. Do not editorialise about your colleagues, and do not present it as a request for someone to take your side.
Two properties make this work. It gives the decision to the person entitled to make it, which is where an appetite for commercial risk properly sits, rather than leaving it to be settled by whoever is most insistent. And it creates a record, which matters because if the risk fires later, the question asked will be whether anybody flagged it. A dated note saying delivery assessed this at a materially higher figure and the decision was taken to submit anyway is not blame-shifting; it is the reason the organisation can learn something instead of re-running the argument.
Then behave well after the decision. If you are overruled and the lower number is submitted, the position to hold is that the assumptions and exclusions in the document match the number that was submitted, because that is the mechanism through which the gap gets recovered legitimately later. Losing the argument about the number and also giving up the protective wording is losing twice.
What delivery receives on day one
The reason this whole question matters is what happens at handover, and interviewers respect a candidate who reaches it unprompted. The estimate is not the deliverable; the reasoning behind it is. Delivery needs the assumptions and their sources, the items priced from evidence versus the items priced from judgement, the exclusions and how they were worded to the customer, the contingency and what it was held against, and every commitment made verbally that never made it into the document.
That last item is the one that is always missing, and it is the same defect as a demo answer nobody wrote down. Something was said in a meeting, the customer believes it, and the team who has to satisfy it has never heard of it. An estimate handed over with its reasoning intact is a plan a team can defend. A number handed over on its own is a number they will be judged against and will not own, which is exactly the position you were trying to avoid at submission.
Nobody should be held to an estimate they never accepted. Get the person who will run the work to own the effort before it leaves, resolve disagreements at the line items rather than at the total, and if the gap stays open, escalate it in writing to whoever is entitled to accept it.
© 2026 Preptima. Originally published at preptima.com (opens in a new tab).