How we price a job.

The short version

The photos and your description determine what the work is. A separate step determines what it costs. Keeping those apart is the whole design: it means we can tell you whether a wrong estimate was a misread of the job or a mispriced material, and it means correcting the scope re-prices instantly without re-analysing anything.

Labor: measured, not guessed

Labor is usually around half the cost of a residential job, and it is the half we can measure. Hourly wages come from the U.S. Bureau of Labor Statistics Occupational Employment and Wage Statistics program, which publishes wagesby trade and by metropolitan area.

With an important limit. BLS withholds figures where a trade has too few employers in an area to publish without identifying them. So a narrow trade in a smaller metro — a tile setter in Bellingham, say — may have no published local rate at all. Where that happens we fall back to the state or national figure, andwe say so on the line item rather than presenting a national average as if it were your local one. That fallback is exactly where we are weakest, so it is labelled, not hidden.

That per-trade detail matters more than it sounds. The spread we measured across U.S. metros runs from about 0.78× the national average to about 1.45× — and it moves differently for different trades in the same city. Washington D.C. electricians run about1.11× national while D.C. general laborers run about0.95×. A single city-wide multiplier, which is what most cost guides apply, is therefore wrong in a way that compounds on labor-heavy work.

A published wage is not what a contractor bills, so we apply a burden and overhead multiplier on top. The wage is measured; that multiplier is our assumption, and we label it as one.

Materials: estimated, and labelled as estimated

There is no material price feed we can honestly offer you. The major construction cost databases are either enterprise-only, do not license their data to third parties, or explicitly forbid this use. So material prices are produced by our models, then adjusted using Bureau of Labor Statistics Producer Price Index series for the specific commodity.

That adjustment is not cosmetic. Between January 2024 and August 2026, gypsum rose about 2% while lumber rose about 21%. A single blanket inflation figure would be nearly twenty points wrong on individual materials.

Every line in your estimate says which it is. Measured data and a model estimate are never presented as if they were the same thing.

Why you always get a range

A single number would be more satisfying and less honest. AACE International, the cost engineering body, classifies an estimate made at this stage of a project as carrying an accuracy of roughly −30% to+50% at its broadest, tightening to about−20% / +30% when the job is well understood.

So our ranges are wide, and they widen further where our confidence is lower. That is the honest output. Research on how people read uncertainty is reassuring here: a numeric range costs almost nothing in credibility, while vague verbal hedging — "roughly", "around", "in the region of" — costs a great deal. A bracket is both more truthful and more trusted than a fuzzy single figure.

The arithmetic

No model ever performs the arithmetic. Models produce quantities, hours and unit prices; deterministic code multiplies, sums, applies waste factors, tax, permits and markup. All money is handled in whole cents, so the line items always sum exactly to the total.

Markup is applied as a divisor rather than a multiplier —cost ÷ (1 − overhead% − profit%). Treating "10 and 10" as a 1.20× multiplier is a common error that yields8.3% of revenue rather than 10%, and it would make every estimate systematically low.

Permits

Permit requirements are set by thousands of separate local authorities, and the exemptions homeowners rely on are usually local amendments rather than the model code. Seattle, for instance, exempts single-family re-roofing entirely. Drywall thresholds vary from 16 square feet to no exemption at all.

So we tell you what is likely, never what is certain, and we always name the authority you should confirm it with. We would rather say we do not know than invent a fee.

Where this method is weakest

• Photographs cannot see through walls. Rot, mould, failed waterproofing and undersized wiring are the things that blow budgets, and they are exactly what a camera misses. We list them as unknowns with their likely cost impact rather than pretending they are not there.
• Material prices are model estimates. Better than a national average because they are adjusted to your market and to current commodity movements — but not the same as a supplier quote.
• Unusual work is harder. Standard jobs in common configurations are where we are strongest. Bespoke or structurally complex work deserves a wider range, and gets one.
• We have no site visit. A good contractor standing in your bathroom knows things we cannot. Our estimate is a reference point for that conversation, not a replacement for it.

What would make us refuse

If our checks reject a figure as implausible, or our confidence falls below the floor, you get no estimate and no charge. A confidently wrong number is worse than no number, because you might act on it.

What we do not do

We take no money from contractors, in any form. No lead generation, no referral fees, no sponsored placement, and your details are never passed to anyone. A price authority that is paid by the people whose prices it describes is not an authority. Keeping that line is the only reason to believe anything on this page.