A construction cost database is only useful when field cost codes map to the estimate’s structure: CSI MasterFormat for trade pricing, UniFormat for building systems and conceptual work. Someone has to reconcile every closed job back to its estimate. Put that seat in estimating, under the chief estimator, and hire an estimator who has closed out jobs.
Two estimators at the same Dallas general contractor price the same 64,000-square-foot medical office building, from the same drawings, out of the same cost database. One lands at $341 per square foot. The other lands at $398. Neither made an arithmetic error, and both can name their comparables. The gap is about $3.6 million on one building (a worked illustration with assumed figures), and it was sitting in the database before either of them opened it. If you run preconstruction, you have sat in that meeting, and you probably settled it by splitting the difference.
Why do two estimators get different numbers from the same cost database?
Because the database stores cost the way the field coded it, not the way the estimate was built. When one job codes ductwork labor to mechanical and the next codes it to a general labor code, the comparables stop being comparable, and each estimator corrects for that from memory.
In the illustration, Estimator A took mechanical from jobs where part of the duct labor had landed in a general labor code, so mechanical looked cheap. Estimator B took the frame from jobs where underground utilities sat inside the concrete code, then priced sitework again on its own. One estimate left cost out. The other counted it twice.
Field cost codes are written for payroll. A superintendent coding time cards on a Friday afternoon needs a code that exists and is close enough. The estimate is built on a different structure, usually two:
| Structure | Organized by | Set by | Good for | Where it breaks |
|---|---|---|---|---|
| Field cost codes | Work as crews perform it | Accounting and operations | Payroll, job cost, the monthly forecast | Codes drift from job to job |
| CSI MasterFormat | Work result and trade | Specification writers | Trade pricing, subcontract scopes, buyout | Hard to use before drawings exist |
| UniFormat II | Building element or system | Estimators and owners | Conceptual estimates, comparing buildings | One trade spreads across several elements |
The federal government settled this for its own buildings long ago. The NIST UniFormat II report says the element structure “ensures consistency in the economic evaluation of building projects over time and from project to project.” GSA P120 requires the ASTM version “for consistency in use and maintenance of GSA’s cost databases,” with CSI MasterFormat listed beside it.
Recoding the five comparable jobs to one structure put the building at $372 per square foot. Estimator A, at $341, was $31 low. Estimator B, at $398, was $26 high. The fault was never in either estimator. It sat in the gap between the people who code cost and the people who use it. Recoding only helps where comparables exist, and a mass timber frame often has none in the history.
Which leaves the question nobody in that preconstruction meeting owns: whose job was it to catch the drift before bid day?
What does the owner of a cost database do?
The owner reconciles every closed job back to its estimate, maintains the map from field cost codes to MasterFormat and UniFormat, adjusts history for time, location and market, and runs the conceptual benchmarks preconstruction relies on before drawings exist. It is an estimating seat that speaks accounting.
Five standing duties make up the seat:
- Closeout reconciliation: final cost against the estimate, by element and by trade, with a written reason for every material variance.
- The code map: one list of field cost codes, mapped to both structures, and the authority to refuse a new code that has no home.
- Normalization: adjusting history for escalation against a yardstick such as ENR cost indexes, and for labor basis. A federal job paid under Davis-Bacon wage determinations, which set locally prevailing wages and fringes, is no comparable for a private job across the street.
- Benchmarks: cost per square foot and per element by building type, refreshed as jobs close, so a conceptual number starts from evidence.
- Feedback: telling the estimator who carried the job where it landed, and why.
The GAO cost guide lists “updating estimates with actual costs” as a step of a reliable estimate. Most contractors stop one step short. A project controls leader owns the forecast on live jobs; this seat owns the finished ones, and the conceptual estimator is its heaviest customer.
The obvious objection is that software now does this.
Estimating software prices faster from the same cost database
Estimating AI is solving the fast part of the job. Pricing was never the bottleneck in preconstruction; scope definition was, and the tools do not touch it. A takeoff tool that reads a drawing set in minutes still prices those quantities against the history it is given. Feed it miscoded jobs and it produces the wrong number sooner, to two decimal places.
The estimate classes in GSA P120, drawn from ASTM E2516 and AACE recommended practices, say it directly: an estimate’s accuracy “is directly related to the level of scope definition.” A screening estimate at 0 to 2 percent project definition is a scope problem first. No software defines scope for an owner who has not decided what the building is.
A technology director can buy the tool. Someone still has to own what it reads, and what it reads gets written at closeout.
Closeout is where the cost history gets written
Closeout is the one moment a contractor knows what a building cost, and it is the moment the people who know why have already moved on. The project manager is on the next job. The superintendent is mobilizing somewhere else. Accounting closes the books and files the final cost report for a purpose that has nothing to do with the next bid.
The split runs through the whole job. The people who answer for the money mostly see the job through reports, and the people who spend it rarely see a cost code. Work done on a verbal instruction lands in whatever code was open that week. Estimating, operations and accounting each hold a third of the explanation, and nobody puts the three together before the team disperses. The closeout leader chases the retention. Nobody chases the history.
The consequence is not abstract. The next bid on that building type starts from a number nobody can explain, it goes out under your signature, and the variance that would have corrected it is in a filing cabinet, coded to the wrong line. Every unreconciled job makes the database a little more confident and a little less right.
Heavy civil carries the same problem in unit costs. FHWA’s estimate guidelines describe historical bid-based estimating as matching line-item quantities to historical unit bid prices, adjusted for location, size and market conditions. The TxDOT estimating manual warns that average low bid prices “should only be used as a starting point.” FHWA’s highway cost index is built from state postings of winning bids. Bid tabs tell a Houston contractor what the market paid for a cubic yard of excavation. Only its own reconciled production rates tell it what that yard cost to move.
Who should own the cost database, and where should the seat report?
Estimating should own it, through a cost data manager or estimating systems manager reporting to the chief estimator or the VP of preconstruction, with a dotted line to accounting for the cost code structure. The person who uses the number has to own its definition, or nobody does.
Accounting owns the ledger, which has to balance; a database has to compare. IT owns the system, not the meaning. Project controls owns live work. A warranty manager, where a builder has one, owns the cost that lands after handover, and that belongs in the history too. When the seat reports to the chief estimator, the person who signs the bid can also say which history counts.
The seat is not for every contractor. A firm that prices hard-bid work mostly from subcontractor quotes, with little self-perform, gets less from its own history than one carrying concrete crews or negotiated work. A firm with one chief estimator and one building type may need a disciplined closeout form, not a hire.
What should you hire for in a cost data manager?
Hire an estimator who has closed out jobs, not a database administrator. The right candidate has mapped cost codes to MasterFormat or UniFormat, reconciled final cost against an estimate, and can ask a superintendent why a code moved without starting an argument.
One interview exercise does most of the work. Hand the candidate three closed-job cost reports and their original estimates, and ask which one is a fair comparable for a new building. A strong candidate asks how the codes were used before touching the numbers.
The pool is narrow. BLS counts 224,220 cost estimators across all industries in May 2025, at a median of $78,740, a figure covering every level of the occupation, not a senior seat. Construction managers sit at a median of $114,990 on the same all-industry basis. Construction employment stood at 8,364,000 in September 2026 in BLS payroll data, and the industry quits rate in BLS JOLTS was 2.2 percent in August. The person who has done this work in Atlanta or Nashville is employed and not reading postings.
Start at our employers page and we will tell you honestly whether your cost database needs a dedicated owner or a firmer closeout routine. Selah Talent Partners runs confidential executive search for mandates that cannot be publicly posted, including estimating leadership a contractor would rather not advertise. The test is whether your newest estimator, handed the same closed jobs as your chief estimator, comes back with the same number.
Sources and further reading
- Cost estimators, BLS OEWS May 2025, all industries
- Construction managers, BLS OEWS May 2025, all industries
- BLS JOLTS, construction quits and openings rates
- BLS payroll data, Current Employment Statistics
- UniFormat II report, NIST, 1999
- GSA P120, PBS cost and schedule management policy
- GAO cost guide, GAO-20-195G cost estimating and assessment guide
- Estimate guidelines, FHWA engineer’s estimate and bid review guidance
- Highway cost index, FHWA NHCCI methodology
- TxDOT estimating manual, estimated bid price determination
- ENR cost indexes, construction and building cost index history
- Wage determinations, SAM.gov Davis-Bacon wage determinations
- Locally prevailing wages, US Department of Labor, Davis-Bacon requirements
- AACE recommended practices, AACE International