Most construction “pricing data” is one of two things: a national cost book that smooths the entire country into a single number, or an estimating model that guesses. Caliche is neither. It is built from real bids that real Texas contractors submitted at public bid openings — the closest thing the industry has to ground truth. This page is the short version of why that matters and how we keep it trustworthy.
1.6 million+
Real bid observations
250+
Texas counties
80,000+
Line items priced
We tested it blind — here's where it landed
Data is only worth anything if it prices a real bid well. So we tested it the honest way: we took 1,870 real Texas bids the engine had never seen, stripped each job's own prices out of the pool first so it couldn't cheat (leak-free), priced every line cold, and measured where the total landed against the actual winning bid.
Two honest things about that picture. First, ~1% is Houston — our deepest market — and statewide is ~3%. That gap is the whole point: the more bids we hold in a place, the tighter we get, so our most-covered market is our sharpest. Second, those are medians — the center, not a promise on any single job. Per-job results spread −17% to +32% against the winner, and we'd rather show you that range than hide behind an average. The winner is the lowest of every bidder on a job — an aggressive number — and landing within a few percent of it, blind, on thousands of jobs, is the bar that matters.
Same inputs, same number — every run. The engine is deterministic: it doesn't roll dice, so the price you see today is the price you'd get tomorrow. And the whole test is reproducible from the public bid record — it isn't a one-time figure we typed in and can't back up.
It's the market, not a model
Every number is anchored to a price a contractor actually put on paper and submitted in public — the real Texas bid record. We don't blend in other states, we don't publish a manufacturer's list price, and we don't ask a model to invent a number. When you see a unit price in Caliche, it's what your competitors actually bid, not what a book three years out of date thinks it should cost.
Depth is the whole point
One observation is one priced line item from one bidder on one project. We count every bidder's submission, not just the winner's — which is why the corpus is measured in the millions and spans 250+ counties, every major Texas region, and 2,000+ public agencies. That breadth is what lets Caliche price your scope, in your market, instead of handing you a statewide average that misleads on both ends.
Every figure carries its sample size. A price backed by 400 bids and one backed by 6 are not the same confidence, and we never pretend they are — you always see the n behind the number, so you can weigh it yourself.
Kept clean, kept current
Real-world records are messy — the same item gets written a dozen ways and the same firm shows up under several names. We do the unglamorous work of reconciling all of it so a single project is never double-counted and a price reflects genuinely comparable work. And the index rebuilds nightly, so newly published tabs are folded in within a day of ingestion rather than waiting on the next annual release.
What Caliche does not do
It does not hand you the winning number, because no one can predict it — even the strongest Texas contractors win a small fraction of what they bid. Caliche shows the observed range your competitors actually submitted for your scope, with the confidence behind it, so you can place your bid with eyes open. It positions; it does not predict. To see how that plays out on a real bid, read TxDOT unit prices in Texas and how agency type drives Houston pricing.
See it on your own scope
Price a bid against the same real Texas market data described here.
Frequently asked questions
Is this real bid data, or an estimate?
Real. Every figure traces to an actual bid a contractor submitted at a public bid opening in Texas — not a national average, not a cost-book line, not a modeled guess. That’s the difference between pricing against the market and pricing against an assumption.
What counts as a "bid observation"?
One priced line item from one bidder on one project — for example, a single contractor’s unit price for 24-inch RCP on a specific job. A bid with 30 items and 6 bidders contributes up to 180 observations. We count real bidder submissions, which is why the corpus runs into the millions and reflects what the whole market bid, not just the winner.
How current is the data?
The index rebuilds nightly, so newly published public bid tabs appear in Caliche within a day of ingestion — not at a once-a-year cost-book release. One honest caveat: the underlying signal moves at the agencies’ pace. Bid tabs post on each agency’s schedule, sometimes weeks after a letting, so you’re seeing the market as current as the public record itself is — refreshed daily, not faster than the record exists.
How accurate is the pricing, really?
We measured it blind on 1,870 real Texas bids the engine had never seen (each job’s own prices removed first). In Houston — our deepest data — the median priced total landed about 1% over the actual winning bid; statewide about 3%. The typical competing bid in those fields ran about 11% over the winner, so landing at 1–3% is competitive 2nd-to-3rd place, blind. Important: those are medians — per-job results spread −17% to +32% against the winner. We show the range rather than hide behind the average, and items we can’t confidently price are flagged Limited rather than guessed.
Does Caliche predict the winning bid?
No — no one can. Even the strongest Texas contractors win a small fraction of what they bid. Caliche shows the observed range real contractors submitted for your scope in your market, with the sample size behind every figure, so you can place your number with eyes open. It positions; it does not predict.