Owners sometimes ask for a peer review expecting a parallel estimate. That is a different, more expensive product. A review checks whether the estimate in front of us is fit for the decision it is meant to support, and says so in writing.
The request usually arrives at a pressure point: a board paper is three weeks out, a lender wants independent comfort, or a bid has produced a number nobody can explain. At that moment the owner does not need a second opinion on every unit rate. They need to know whether the number can carry the weight about to be placed on it, and if not, where it will crack.
Review, validation, or a second estimate
Three different products get called “a review”, and buying the wrong one wastes money in both directions. A full independent estimate rebuilds the number from the documents: strongest evidence, highest cost, weeks of effort. A validation tests the estimate from outside: benchmarks, metrics, high-level checks. A peer review goes through the estimate itself: its basis, structure, quantities, pricing and risk, without rebuilding it. The discipline is described in AACE International Recommended Practice 31R-03, Reviewing, Validating, and Documenting the Estimate (RP 31R-03), available to AACE members at web.aacei.org. The review is usually the right purchase: deep enough to find the real problems, fast enough to fix them before the decision.
The eleven questions
Every review we run asks the same eleven questions of someone else’s number.
- Does the basis of estimate match the current scope documents?
- Is the class claimed supported by the maturity of the inputs?
- Are quantities traceable to drawings, with allowances shown separately?
- Are the unit rates built up, and for this location and date?
- Are indirects schedule-based?
- Is escalation shown separately, by commodity, to the spend midpoint?
- Is the contingency derived from a risk analysis, and does the register exist?
- Are owner’s costs included, and does the owner agree with what is in them?
- Are exclusions listed, and is anything in them actually required?
- Does the estimate reconcile to the previous class?
- Does it benchmark inside a normalised range?
The two that matter most
Questions one and two. Almost every estimate that fails in construction failed here: it described a project that had already changed, or claimed a precision the inputs could not support.
The first failure is quiet. The design moved, the battery limits shifted, a scope decision was taken in a meeting, and the estimate kept pricing the older project. Nothing in the arithmetic is wrong; the subject is. The second failure wears a label: a “Class 3” whose quantities are factored and whose indirects are a percentage is a Class 5 in a nicer binder, and the accuracy range everyone is relying on is fiction. What separates the classes, really, is its own subject: see What actually changes between a Class 5 and a Class 3 estimate.
The remaining nine questions matter, and any of them can produce the biggest finding on a given job. But one and two decide whether the estimate is even about the right project at the right maturity, which is why the review starts there and spends a third of its time there.
How the review runs
The request list is short: the estimate file, the basis of estimate, the WBS, the risk register, the current scope documents, and the rate build-ups for the significant accounts. A review that cannot get these documents has already produced its first finding.
The method is sampling with intent, not full re-measurement. The cost is concentrated: on most estimates, twenty accounts carry eighty percent of the money, so the review drills those. A handful of quantities get traced from estimate line back to drawing and revision. A few rates get walked through their build-up. The indirects get tested against the schedule, the escalation against the buy dates, the contingency against the register. Where a thread pulls loose, the review follows it; where everything traces, it moves on.
Findings come back ranked by cost and risk significance, each with what it is, what it is worth, and what to do about it. Ten findings in priority order get fixed. Forty findings in alphabetical order get filed.
What findings look like
From an illustrative review, the top three might read like this. First: indirects are carried at 22 percent of directs on a 30-month remote project; a schedule-based build-up indicates 15 to 20 million more, and this is the largest single exposure in the estimate. Second: escalation is priced to the pricing date only; the project spends over three years, and the missing escalation is worth roughly four percent of base. Third: earthworks quantities cannot be traced to the current grading plan; the take-off references a superseded revision, and the delta is unquantified until re-measured.
Notice what the ranking does. The owner now knows the number is more likely light than heavy, by roughly how much, and which three actions close most of the gap. That is what “fit for the decision” means in practice: not a perfect estimate, but a known one.
A review that only checks the arithmetic has checked the part that was never wrong.
Where this goes wrong
The review commissioned too late. The board meets Friday; the review lands Thursday night. Whatever it finds, there is no time to fix, so the findings become footnotes to a decision already made. A review needs three to six weeks of runway to be worth its fee.
The friendly reviewer. The review is done by the same firm that built the estimate, or by a team that reports to the project it is reviewing. The independence is organisational theatre, and lenders in particular can smell it. Independent means a different firm with nothing at stake in the answer.
The arithmetic audit. The review checks that the spreadsheet adds up and the rates match the database, and pronounces the estimate sound. Spreadsheets almost always add up. The failures live in the basis, the maturity, and the assumptions, which an arithmetic audit never opens.
The unranked list. Every observation from material to trivial, presented flat. The project team fixes the easy ones, declares the review addressed, and the two findings that mattered are still in the estimate at sanction.
The adversarial frame. The estimating team treats the review as an attack and defends every line. The reviewer is not the opponent; the tender results are. A review finding caught in week three costs a memo. The same finding caught by the market costs the margin.
What the owner receives
A written opinion the board can rely on: whether the estimate is fit for the decision, the findings ranked by significance, and clear recommendations for correction or further investigation. Sometimes the opinion is that the estimate is sound, and that sentence, from an independent reviewer, is worth the fee on its own, because it converts a number into a defensible number.
Emerald Group performs independent estimate reviews for owners, engineering firms and lenders, ahead of board papers, sanctions and bids. If a number is about to carry a decision and nobody outside the project has read it, get in touch.