How Permit Budget estimates work

What the numbers are built from

Every estimate starts from actual permit filings. When a contractor pulls a permit, the filing includes a declared job valuation: what the applicant told the buildings department the work would cost. We collect those valuations by trade and by market, thousands of them, and build our estimates from the distribution of what was actually filed. Not national averages. Not a rule of thumb. What permits in your market said, trade by trade.

What we do with the filings

• We pull initial permits only. Renewals and amendments are tracked for volume but never enter the cost statistics, because a renewal is paperwork, not a new price signal.
• We map each permit's work type to a trade (plumbing, HVAC, general construction, and the rest) using a versioned mapping we publish below.
• We aggregate by market, trade, and month: the 25th percentile, the median, and the 75th percentile of filed valuations, plus how many permits sit behind each number. You see the sample size with every estimate.
• We do not serve a number unless at least 30 permits back it. Thin data gets an honest "not enough permit data yet," never a guess.

The understatement problem, stated plainly

Permit valuations understate true contract cost. Filers have reasons to file low, and everybody in construction knows it. We do not pretend otherwise. Our normalization rules adjust for this bias per trade, and every adjustment is earned: a multiplier only ships after blind tests against real bid outcomes prove it improves accuracy. The current normalization version is shown with every estimate, and the version history is public on this page.

What an estimate is not

A Permit Budget estimate is a planning-grade number grounded in permit filings. It is not a bid, not a quote, and not a substitute for pricing from a licensed contractor who has walked the job. Use it to sanity-check a bid, to underwrite a deal, to budget a capital plan. Then get real numbers before you sign anything.

Known limits

• Floor area: permit records do not include square footage, so our trade bands are whole-scope ranges unless stated otherwise. Where we show per-square-foot figures, they are modeled and labeled as modeled.
• Electrical and elevator work: in New York these are filed through separate channels outside our current dataset, so we do not estimate them yet. We say which trades are covered before you pay, and we will not sell you an estimate for a trade we cannot back with data.
• Geography: estimates only cover markets that have passed our blind-test accuracy gate. A market with thin coverage stays closed.

How we check our own work

Before any market opens, and monthly after, we run blind tests: real project scopes priced by our engine and by generic calculators, scored against actual bid outcomes. A market opens only when our estimates beat the generic alternatives. The latest results are published below, win or lose.

Normalization versions

• v0 (current): raw filed valuations, no adjustment. The bias is documented here instead of hidden.
• v1: pending blind-test evidence. No multiplier ships without it.

Work type to trade mapping, New York City (v1)

Scaffolds, sheds, fences, signs, and similar site logistics permits are collected but excluded from estimates.