Final bid review: a QA/QC playbook for general contractors

A final bid review is the controlled decision gate between a working estimate and a price the company is prepared to submit, explain, contract around, and ultimately build. It is not a math check. Arithmetic is the easiest part of it and the part least likely to lose money.

The review confirms something harder: that the correct documents and addenda were used, that the complete project scope has been assigned, priced, excluded, qualified, or carried as risk, that subcontractor bids were normalized without hiding uncertainty, that self-perform quantities and general conditions and escalation and fees are internally consistent, that the estimate reconciles to history and to an independent perspective, that the submission complies with the owner's instructions, and that every unresolved risk has a named owner and an approved financial treatment.

The rule underneath all of it is short:

A bid is not ready because the spreadsheet totals. It is ready when the price, scope, assumptions, risks, documents, and submission requirements tell the same story.

Six tests a finished bid has to pass

AACE International describes estimate review as a structured process that should confirm scope coverage, compliance with required practices, freedom from material errors and omissions, and suitability for the estimate's intended purpose. The GAO cost-estimating framework adds four characteristics of a reliable estimate: comprehensive, well documented, accurate, and credible. Translated into a general contractor's closeout week, they become six practical tests.

A bid can pass five of these and still be the one that hurts you. The compliance test is the clearest example: an estimate can be technically excellent and still be rejected because an alternate landed on the wrong bid-form line.

What to call it internally

The terminology matters more than it sounds, because the name determines when people think the work starts.

Use final bid review as the name and describe it internally as the company's pre-submission estimate QA/QC gate. Then fight the word "final," because the single most common failure in this process is starting it on bid day.

It is broader than bid leveling

Bid leveling compares subcontractor proposals on a common scope and commercial basis. Final bid review contains leveling and then keeps going, because most of the money that escapes an estimate is not inside the subcontractor columns.

A low subcontractor number can still produce an unsafe GC bid if the estimate is missing temporary works, schedule impact, design risk, or a compliant bid form.

Set the standard before bid week

The review should be defined in the estimate plan, not improvised in the final hours. AACE recommends that reviews be planned and managed by a designated lead, include the right estimating, scheduling, procurement, technical, execution, and management participants, and record action items throughout, and, critically, that the estimating schedule reserve enough time for findings to be corrected rather than discovered after the price is effectively locked.

Five things should be settled at estimate kickoff.

An example closeout week

The shape below is a medium-to-large hard bid. Scale it to package count, delivery method, and design maturity, but keep the property that matters: findings surface with time left to price them.

Anything discovered at T-0 gets priced with a round number, because there is no time for anything else. That is the argument for the calendar.

The workflow

The sequence holds across delivery methods. Depth is what changes.

1. Confirm bid requirements and freeze the document baseline.
2. Freeze the estimate structure and the scope map.
3. Review trade coverage and self-perform scope.
4. Normalize subcontractor and supplier proposals.
5. Review quantities, rates, extensions, and estimate formulas.
6. Review general conditions, schedule, logistics, and procurement.
7. Identify scope gaps, commercial issues, and design risks.
8. Reconcile against history, budget, and an independent check.
9. Set carry numbers and risk treatment.
10. Complete operations and management review.
11. Verify bid forms, addenda, attachments, and submission mechanics.
12. Lock, submit, and archive the approved estimate.

Start with the document baseline

Before checking a single estimate cell, confirm the source set. Everything downstream inherits its errors.

Federal sealed bidding is instructive even where it does not govern. The rules distinguish responsiveness from responsibility and address apparent mistakes and bid rejection separately, which encodes a lesson that applies to any procurement: a materially noncompliant bid cannot always be repaired with a spreadsheet adjustment. Private subcontractor leveling has more latitude, but the same instinct protects you.

Build a source-backed scope map

The core review artifact is a scope map, not a cost-code report. Each requirement should be traceable along one chain:

Source requirement → scope item → responsible package → bidder response → estimate line → carry treatment → open issue

AACE's basis-of-estimate guidance asks for the same content in narrative form: project scope, pricing basis, allowances, assumptions, exclusions, risks, opportunities, deviations from standard practice, and key agreements. The map is what makes the narrative auditable.

At minimum, cover Division 00 and 01 requirements, permanent systems, temporary works and utilities, access and hoisting and protection, testing and commissioning and training, mockups and delegated design, BIM and coordination, permits and fees and bonds and taxes and freight, warranties and closeout and attic stock, phasing and premium time and shutdowns, owner-furnished items, design-assist and deferred submittals, demolition and patching, and every interface between adjacent packages.

Division 01 is the one teams skip. It is also where temporary, administrative, quality, and closeout costs live, and none of them appear in a trade proposal.

Run separate technical, commercial, and compliance lanes

Combining every issue into one meeting makes it impossible to tell a scope problem from a submission problem, and the loudest issue wins rather than the largest one.

The pre-submission checklist

Normalize scope before price

The goal of normalization is not to make every proposal look equal. It is to expose the economic and operational consequences of each bidder's actual offer. That requires keeping three values separate and visible.

Never overwrite the submitted number. The original proposal should stay visible and linked to its source, because three weeks later an estimator adjustment reads exactly like a subcontractor commitment unless the sheet says otherwise.

Four statuses, and only one of them is silence

Silence is not inclusion. It may eventually be confirmed as included, but until then the estimate is relying on an assumption nobody recorded and nobody agreed to. Do not mark an item included because a different bidder included it.

Choosing the right adjustment

The full method for turning these adjustments into a number you can defend (the evaluated-cost formula, expected-value risk pricing, and the duplicate-contingency trap) is covered in how to set a carry number you can defend.

Gaps and overlaps are different problems

A gap exists when required work has no clear owner. An overlap exists when more than one package carries the same work. Both cost money, in opposite directions: a gap becomes a change order or a margin loss, while an overlap inflates the bid, obscures the true low bidder, and quietly disappears during negotiation when both trades assume the other stayed responsible.

Interfaces deserve their own pass. Firestopping, cutting and patching, equipment controls, supports, curbs, sleeves, access panels, temporary protection, power wiring, low-voltage wiring, controls integration, excavation and backfill, housekeeping pads, startup, and closeout are the usual suspects. Scope gap review has its own sequence and is worth running as a distinct step rather than folding into leveling.

Prioritize findings before you chase them

Not every open item deserves a clarification two days before close. A dollar threshold alone is the wrong filter: a small compliance error can invalidate a public bid, while a larger cost issue may be a transparent owner allowance.

A gap log entry has to be specific enough to act on. "Check firestopping" is not a finding. "Confirm furnishing and installation of firestopping required by Spec 07 84 00 at all mechanical, electrical, plumbing, and fire-protection penetrations" is a finding, because someone can answer it.

The hard cases

Explicit exclusion. Price the excluded requirement from the best available source and seek confirmation. If the exclusion shifts contractual responsibility or means and methods, it is not just a price plug.

Scope silence. Ask the bidder directly and record the item as unresolved until you get an answer in writing.

Bundled bid. Compare the bundle as a bundle. Allocating arbitrary values to individual trades and then treating those values as the bidder's submitted numbers invents a proposal nobody made. Bundling can be an advantage because it removes interfaces, or a risk because it concentrates exposure, assess the net effect rather than the arithmetic.

Very low bid. Do not assume it is wrong, but review the quantity basis, addenda, exclusions and silence, labor and material coverage, shifts, insurance and bonding, tax and freight, supplier quotes, subcontracted portions, capacity and workload, and the possibility of an obvious clerical error. Federal procurement treats apparent clerical mistakes differently from strategic pricing for a reason.

One bid only. A single bid is a price, not market validation. Test it against an independent estimate, historical cost, current benchmarks, a comparable package on another live project, supplier-level quotes, or a self-perform analysis.

No conforming bid. Do not select a nonconforming proposal because it is the only number. Carry an independent estimate, rebid, resplit the packages, negotiate with a qualified partner, carry an allowance, defer award, self-perform, or request a bid-date extension where permitted.

Fifteen questions before you sign

The lead estimator should be able to answer all of these without opening a file.

1. What are the five largest cost assumptions?
2. What are the five largest unresolved exposures?
3. Which packages have weak or single-source coverage?
4. Which selected bids changed materially after normalization?
5. What scope is carried outside the trade packages?
6. Which requirements rest on design interpretation rather than explicit documentation?
7. What would operations challenge first?
8. Which late addenda changed price, scope, or schedule?
9. What is not included in the submitted price?
10. Where could the estimate be double-counting scope or risk?
11. Which quantities or rates would most change the result if they were wrong?
12. Can every material adjustment be traced to a source?
13. Does the estimate reconcile to the owner's bid form?
14. Is the proposed margin appropriate for the remaining risk?
15. Would the team make the same decision if bidder names and submitted totals were hidden?

The last one is the test that catches the failure mode nobody admits to: familiarity, reputation, or an attractive low number overriding the evidence in front of everyone.

The review changes with delivery method

A universal checklist is a baseline. It does not replace delivery-method judgment, because the commercial environment decides which failure is most likely.

Design-bid-build

Construction documents are generally complete before bids are requested, which moves the risk from design uncertainty to execution mechanics. On a public school bidding base bid plus three alternates, unit prices, addenda acknowledgement, a bid bond, and subcontractor listing, the review should prioritize exact bid-form compliance, addenda acknowledgement, alternates cleanly separated from base scope, required signatures and bonds, reconciliation between the internal estimate and the owner's form, late trade coverage, and a time-stamped submission receipt.

The common failure is not the estimate. It is an alternate entered on the wrong line, or an addendum that was priced but never acknowledged. The mitigation is procedural: one person prepares the bid form, a different person compares it line by line against the approved estimate and the instructions.

CM-at-risk

Pricing may be established while parts of the design are incomplete, so the review shifts from completeness to maturity. Approaching a GMP at 75% design with interiors and controls unfinished, prioritize reconciliation from the last owner estimate, design maturity by system, what remains inside allowances, buyout status, contractor versus owner contingency, package gaps and overlaps, escalation and quote validity, long-lead releases, design-assist assumptions, and the exception log.

The common failure is a general "design contingency" carried without naming the incomplete systems. During buyout, every piece of scope growth gets argued to be inside the GMP, and there is no record to argue back with. Maintain a design-maturity register that maps each incomplete requirement to an allowance, a contingency, an exclusion, an owner decision, or a defined contractor responsibility.

Design-build

The design-builder carries single-point responsibility for design and construction, so requirement compliance sits alongside trade pricing. Prioritize owner criteria and proposal compliance, design narrative versus estimate scope, delegated design responsibility, performance assumptions, quantity development from incomplete design, code and permitting assumptions, design fees and consultant scope, approvals and design production in the schedule, deviations, and the risk that the solution changes after award.

The common failure is a design narrative promising a system or performance level that never made it into the estimate. Run a requirement-to-design-to-estimate pass: every material owner criterion should be visible in all three.

Negotiated

Repeated budget revisions create assumption drift, and drift is invisible without a cumulative log. An item removed in one estimate returns in the design and not in the cost, and nobody notices until buyout. Keep a running reconciliation that records every addition, deletion, transfer, allowance change, and owner decision since the previous estimate, that log is also the entire basis of owner trust in an open-book model.

Review emphasis by project stage

Who owns the decision

The estimate lead coordinates. No single estimator should be expected to approve every technical, operational, commercial, and executive question.

An approval matrix removes the ambiguity that shows up at 4pm on bid day. Normal leveling adjustments are recommended by the estimate lead and approved by the chief estimator. A major scope gap adds operations approval of the execution plan. An unusual contract qualification is led by risk or legal with executive approval of material exposure. Changes to margin or contingency are executive decisions. A nonconforming public bid issue goes to legal, not to the spreadsheet.

What automation can and cannot own

General-purpose AI is genuinely useful for drafting clarification questions, summarizing a clearly identified document, suggesting checklist categories, structuring a table, and helping an estimator search a limited noncritical document set.

It should not be the sole control for determining whether a requirement is covered across drawings, specifications, addenda, scopes, and proposals; identifying every scope gap; interpreting an unusual contractual obligation; setting a carry number; approving a risk adjustment; determining responsiveness; or replacing operations and management signoff.

NIST identifies confabulation (plausible but inaccurate output) as a material generative-AI risk, and recommends evaluating systems in their intended operational setting, measuring erroneous output, verifying sources and citations, documenting limitations, involving domain experts, and retaining human review. That is the right posture here, and the reason is not only that a model can write a wrong sentence. A final bid review needs controlled access to the correct revisions, construction-specific interpretation, company scope conventions, historical data, permission controls, and a record of what a human accepted.

How to validate a system before you trust it

Do not start with a demo question like "what are the project risks?" Build a controlled test set from completed projects instead.

1. Select representative trades, delivery methods, and document conditions.
2. Identify the known inclusions, exclusions, alternates, silences, addenda changes, and estimate issues.
3. Have senior estimators create the approved answer set.
4. Run the system without revealing the expected answers.
5. Score extraction accuracy, issue recall, false positives, source correctness, and reviewer time.
6. Separate "found the text" from "understood the construction implication."
7. Measure by trade and document type, not as one aggregate score.
8. Repeat after any model, prompt, template, or data change.
9. Require human approval before anything updates the estimate.
10. Keep the results for governance and vendor management.

What to measure

There is no authoritative industry benchmark for "percentage of final bids containing an estimating error," and any article that gives you one invented it. Companies define errors, estimates, stages, and impacts differently. Adjacent benchmarks still make the case: the Construction Industry Institute reports rework ranging from 2% to 20% of contract value, and an Autodesk and FMI study attributed 14% of 2020 rework to bad data, with 30% of respondents calling more than half of their project data bad. Neither figure is an estimating error rate. Both describe the cost of information that was incomplete, incorrect, or late.

Which is the argument for measuring your own.

Do not reward a system for producing more issues. Reward it for finding material issues accurately, with evidence, early enough for the team to act.

Where teams get this wrong

Where Piper fits

Final bid review is judgment work sitting on top of an enormous amount of reading, performed under the worst time pressure in the estimating calendar. The judgment is what estimators are paid for. The reading is what runs out of hours, and it is the first thing to get compressed when three pursuits peak in the same week.

Piper brings subcontractor proposals, drawings, specifications, addenda, and your company's scope standards into one source-linked review. Estimators start from scope rows derived from the actual bid set, see exclusions and silent items surfaced with the clause that produced them, apply their own adjustments, and keep the decisions that need to reach the estimate and the subcontract. The review still belongs to the estimator. The hours spent finding what to review do not have to.

Sources

• AACE International, estimate review and basis-of-estimate guidance
• U.S. Government Accountability Office, cost-estimating framework
• Construction Industry Institute, rework ranges as cited in the body
• Autodesk and FMI, rework and project-data quality findings as cited in the body
• Construction Management Association of America (CMAA), delivery-method guidance
• Design-Build Institute of America (DBIA), delivery-method guidance
• National Institute of Standards and Technology (NIST), generative-AI risk guidance