About Hubstake
A better way to understand
what construction costs.
Public agencies already publish what they paid — as tens of thousands of PDFs, in a dozen incompatible layouts, with no search across them. Hubstake reads those documents and turns them into something you can look a price up in, without breaking the link back to the page it came from.
Transparency
Every figure carries a link to the public document it was read from. No proprietary adjustment, no blended index, no black box — if you disagree with a number you can open the document and say why.
Coverage
3,916,866 observations from 64,484 public documents, published by 19 agencies across 18 states. The coverage page is generated from the index on every build, so it cannot claim more than the site serves.
Accuracy
Quantity × unit price has to reconcile against the extended amount printed on the document before a row is published. Rows that do not reconcile are quarantined, not rounded into place.
Built for professionals
For estimators, owners’ representatives, procurement teams and infrastructure investors — people who need a defensible number rather than a ballpark.
Why this exists
Construction pricing is one of the few markets where the transaction record is public and almost nobody can use it. Every state transport department publishes bid tabulations — the unit price every bidder offered for every pay item, and which bid won. That is a real, dated, sourced price for real work, and there are millions of them sitting in public document repositories.
In practice they are unusable. They are scanned or fixed-width PDFs, each agency in its own layout, with no index across agencies and no consistent naming: the same pipe is “RC PIPE (CL III)(24 IN)” in one state and “24" REINFORCED CONCRETE PIPE” in the next. Answering “what has 24-inch RCP been going for?” means opening documents one at a time.
So the alternative most people reach for is a subscription index that returns a national average with no way to see what it was built from. That is a number you cannot defend in a negotiation, because you cannot show anyone where it came from.
What we do instead
We read the documents, extract every priced line, check the arithmetic against what the document itself prints, group descriptions that mean the same work, and publish a median and a p25–p75 range — with a link to the source document on every single row behind it.
Where we cannot do that honestly, the page says so. 170,679 of 251,133 items rest on a single observation, and those pages show no median and no range, because one contractor’s price on one job is not a market rate. Awarded prices, submitted bids and final payments are three different facts and are never mixed into one figure. Losing bids are read by the parsers and deliberately not published: an offer nobody accepted is not a price.
The methodology page sets out every step, every check and every known limitation in full.
Where the data comes from
Only records an agency has already posted online. No public-records requests, no vendor registrations, no private or commercial sources, nothing behind a login. Where a site asks automated clients to stay out, we stay out and record it as such rather than working around it.
Status
Hubstake is in beta and coverage is expanding. The index currently holds 3,916,866 observations spanning Jan 27, 1998 to Sep 22, 2026. If there is an agency or a jurisdiction you need, tell us — coverage requests genuinely change what gets built next.
Look up a price and check it yourself.
Search is free, and every result opens the document behind it.