How a review works

ClearBid is a second read on a construction bid. It reads what you were given, works out what each bid actually includes, and tells you what is missing, unclear, or in conflict — before you sign something.

This page describes how that is done, and what it deliberately does not do. If you are holding a report and wondering how it reached a particular conclusion, this is the honest version of the answer.

A bid is judged against its own trade

The thing that makes a review worth reading is that it knows what this kind of work goes wrong at. A generic checklist — permits, disposal, cleanup — applies to every project and therefore says nothing about any project.

So a re-roof is held against whether there is a stated unit price per sheet of decking found on removal, whether step and counter flashing are replaced rather than reused, and whether intake and exhaust ventilation are in balance. A shower is held against which waterproofing system is named and whether it is used as a complete system, whether there is a sloped bed beneath the pan liner rather than a flat one, and whether the fan is ducted outside rather than into the attic. A solar array is held against whether the roof has the remaining life to carry it for twenty-five years, and whether the main panel can take it. None of that transfers between trades, and a checklist trying to cover all of them covers none.

Reviews currently carry that depth for:

abatement and remediation · accessibility work · ADUs and conversions · bathroom work · concrete and flatwork · decks and balconies · demolition · driveways and paving · drywall and texture · electrical work · elevators · fencing and gates · finish carpentry and millwork · fire sprinklers and alarm · flooring · framing · gutters and drainage · HVAC work · insulation and air sealing · kitchen work · landscape and irrigation · masonry and chimneys · painting · patio covers and sunrooms · plumbing work · pools and spas · roofing · seismic retrofit · septic and private water · solar and storage · structural work and additions · stucco and plaster · termite and dry rot repair · waterproofing · whole-house renovations · windows and siding

Work outside that list is still reviewed — the same reasoning is applied to whatever trade is in front of it: what gets routinely left out, what is routinely ambiguous, and what is routinely discovered only once work has started.

Alongside the trade, every bid is read for the same four things

• Where money hides. Allowances that are placeholders rather than prices, exclusions (often the most informative part of a bid and the least read), and whether there is an agreed unit price for anything the scope opens up but nobody has inspected.
• Contract and payment structure. Who carries which risk, whether payment is weighted ahead of the work, and whether there is a written process for pricing a change. These are frequently worse than the scope and far less examined.
• Sequence and lead time. What has to happen before what, and which items cannot be compressed no matter who is pushing.
• What applies where the work is. The requirements that are both expensive and routinely absent from bids in a given area.

The numbers are checked, not read

Arithmetic is the one part of a review you can test with the phone in your hand, which makes it the worst possible place to be wrong. So it is not left to judgement: the figures are read off your documents, and the calculation is done in code.

A column of line items is added up and compared against the total printed beneath it. A percentage is checked against the base it claims to be a percentage of. A roof area given in both squares and square feet is converted and compared. If a check does not hold, the finding is removed rather than corrected — and if two figures turn out to agree, nothing is raised at all, because a reconciliation error that is not really there would send you to query a number that was right.

How seriously something is taken is decided in code

Each finding is classified as one of twenty-one named kinds of problem, and the severity follows from the kind rather than from a judgement made in the moment. That exists because the same bid, reviewed twice, would otherwise come back with the same problem called "medium" once and "high" the next time — and a severity that moves is not one anybody can make a decision on.

High means money, a stopped job, or a risk you did not agree to carry. Medium means a likely dispute or change order, recoverable if raised now. Low means worth clarifying for confidence. A high-severity finding is never dropped to keep a report short.

What a review does not do

• It never gives you a price. Not a figure, not a range, not a "typical cost" — not even as a sanity check. Prices move, they are local, and a confident wrong number is worse than no number because it becomes an anchor in a real negotiation. What it does instead is compare the numbers your own bids contain, against each other.
• It does not look up your jurisdiction's code. It can describe what the widely adopted model codes and standard trade practice commonly require, and it names that as the common requirement. Local amendments are real and vary by city as much as by state, so anything in that category is phrased as something to confirm with your building department or a licensed local professional. That is a limit, not a hedge.
• It is not an inspection. It reads documents. It has not seen your building, opened a wall, or measured anything.
• It is not a licensed estimator, engineer, architect, or attorney, and it does not give legal advice. It is there to help you ask better questions.
• It does not keep your bids. Your project details, uploads, and the report itself are not stored. See the Privacy Policy for the one narrow exception, which is the counter described below.

What the bids keep missing

Every finished review adds one to a running count: the kind of problem found, and the kind of work it was. Nothing else — not your bid, not your report, not your account, nothing that could be traced back to you. It exists so this section can say something true about how often each problem actually turns up, rather than asking you to take our word for it.

Not yet. Publishing a percentage drawn from a handful of bids would be the kind of claim that falls apart under one question, so nothing is shown here until at least 50 bids have been reviewed, with at least 20 in any trade before that trade is quoted. The count is running; this section fills itself in.

Seeing it on your own bid

The free preview needs no account and runs this same analysis — it tells you what it checked and what it found, and holds back the detail. It is the shortest way to judge any of the above for yourself rather than taking the page's word for it. Preview a report.

If a review gets something wrong

Tell us. There is a "was this useful?" box at the end of every report, and it goes straight to a person. A review that misreads a bid is worth knowing about more than a review that went fine, and the checklists above are where that correction lands.