RFQ Hunter: 13 Million Award Lines, One Developer, and the Price That Wins

Every weekday before dawn, somewhere in America, a small machine shop powers up. The owner can cut the part, hold the tolerance, and ship on time. Reading ten thousand federal solicitations before the coffee is done is another matter.

RFQ Hunter is my answer to that problem, and it is the largest thing I have built: a government-contracting intelligence platform, built alone, from the ingestion pipelines to the database to the application. The public page is open to anyone. Accounts are another matter — RFQ Hunter is invite-only during the private beta, and an account requires a real government-contracting business.

The situation

The Defense Logistics Agency buys billions of dollars in parts every year, and it posts the work in the open on DIBBS, its bidding board. That was always the cruel joke of this market: the work is right there. It just arrives as a firehose — a typical daily file runs close to two thousand solicitation lines, and the open pile sits above ten thousand at any given time, buried under stock numbers and clauses. Almost none of them are yours.

The big primes handle the firehose with staff and five-figure software. A small shop has a spare hour at night, a browser tab, and a guess at the price that might win. Guess too high, you lose. Guess too low, you bleed. Miss the one RFQ built for exactly what you make, and you never know it existed.

I knew this world before I built for it. RFQ Hunter grew out of client work for a two-person DoD parts supplier — the engagement I write about as ACME Smart Log. Building their operations platform meant learning DLA contracting from the inside, and that client is set to become an RFQ Hunter customer.

What was actually wrong

Not secrecy. The government publishes almost everything: every award and the price paid, stock levels, reorder points, lead times, its own demand forecasts, who is approved to make each part, which weapon system a part belongs to. The problem is the shape. The data sits in flat files, FOIA reading rooms, and portals behind web application firewalls — millions of rows, cryptic codes, no way to ask a normal business question like "what has this part been selling for?"

So the market split. Modern venture-backed GovCon software chases services companies, because text-heavy proposals are the easy application of language models. The industrial base — the machine shops, distributors, and component makers bidding on DLA work every day — was left with legacy tools priced for the primes. The data that would level the field sits published and unread.

That is a data engineering problem. So that is how I treated it.

What I built

The data foundation. RFQ Hunter ingests the government's own published files and makes them usable: 13 million deduplicated DLA award lines covering 2013 to 2026, each with the price paid. 1.25 million stock-level and reorder-point rows. 5.1 million lead-time rows. 5.4 million links from parts to the weapon systems they serve. 3.8 million part-number references connecting manufacturer part numbers to stock numbers to the companies that make them. Nearly 850,000 clear-text specification sheets, and military packaging requirements for over a million parts. None of it is secret. It is public data that was never in a usable shape — that is the moat.

Keeping it current is automated. DIBBS sits behind a web application firewall and a DoD consent banner, so a headless browser clears both on a schedule every morning and pulls the day's solicitations. Awards refresh monthly from DLA's FOIA portal.

Matching. Tell the platform what you make — stock numbers, part numbers, federal supply classes, or your own award history — and the feed narrows ten thousand open solicitations to the short list with your name on it. Each row shows quantity, due date, estimated value, and whether DLA already has stock. Triage is one click: save to the watchlist, hide it, or open a bid. A part-number search works in the direction shops actually think: type the manufacturer's number, land on the government's.

Pricing. For 1.2 million parts, the platform computes a Winning Zone: the band between the 25th and 60th percentile of what the government actually paid, computed over the most recent three years of awards, with the median and the last award beside it. DLA's own price-review flags surface on the same history — how the agency justified each price. The 11 p.m. pricing guess becomes a read of the published record.

Forecasting, as published data. DLA publishes its own 24-month demand forecast for the parts it manages. RFQ Hunter imports it — 2.4 million rows — and puts it beside current stock and reorder points, labeled in the interface as exactly what it is: "DLA published forecast." The platform does not compute predictions of its own. Re-Buy Radar reads the same stock data and flags parts that have fallen below DLA's own reorder point — work likely to be bought again before the solicitation exists.

The competitor view. Every company in the data has a dossier: award totals, how many parts it wins, its product mix, its year-by-year trajectory, and who it actually competes against. Concentration and rivalry measures show whether a niche is sewn up or winnable, and a price-band view shows the real historical prices a rival has been winning at — labeled as history, not prediction. A debarment flag screens the SAM exclusions list by exact CAGE code only, because a false debarment flag is worse than a miss. And the names resolve: of nearly 19,000 companies appearing in the feed, all but two display a real company name instead of a raw code.

The bid workspace. An RFQ opens into a working bid: editable line items, live totals, markup and margin moving together, the Winning Zone band overlaid on the part being priced, and the packaging requirements inline — the classic forgotten cost that eats a margin. The output is a bid document, and then a pipeline: Draft, Submitted, Won or Lost. A won bid flows into delivery tracking through shipped, invoiced, and paid.

Operations around the bid. The platform also runs the shop side of the work: inventory with a movement ledger, jobs that commit stock automatically as they progress, a contacts and subcontractor directory, a compliance calendar that reads live SAM registration data, and a contract book. A supplier can run its day here, not just find work here.

What changed

• Live in production at rfqhunter.com, in private beta with real users; the first real subscriber onboarded against live production data in June 2026
• More than 400 merged pull requests, all by one developer
• 13 million award lines — plus millions of rows of stock, forecast, lead-time, and part-reference data — behind every screen
• Winning Zone pricing computed for 1.2 million parts, verified by hand calculation
• Daily automated ingestion from DIBBS; monthly award refreshes from DLA's FOIA portal
• Free during the beta by design; payment plumbing exists but is deliberately not switched on

What this says about how I work

Numbers get verified before they ship. Every data phase ends the same way: pick a real part and recompute its stored statistics independently — one reference part carries nine awards, a 25th percentile of $93.86, a 60th of $127.98, and the stored values matched every figure exactly. Then trace a stored row back to the raw government file, field by field. Then check the screens against a real supplier's account — the standing benchmark tenant matches 430 live RFQs — because regressions show up in real data that test fixtures hide.

Honesty is built in as a feature, not a disclaimer. The forecast is labeled as DLA's because it is DLA's. An award with no captured price shows a dash, never a fake $0.00. A code the documentation cannot decode is shown raw, never invented. There is a "Data & limitations" page inside the product that says where every number comes from, and the fine print across the app renders from one registry so it cannot drift from the truth.

And scope holds. One person built the ingestion pipelines, the database, the intelligence layer, and the application — because the alternative, coordinating all of that across a team I did not have, was never on the table. The constraint became the architecture: one codebase, one source of truth, every claim traceable to a government file.

The name is the thesis. In this market, the work does not come to you — you go get it. RFQ Hunter reads the field, tunes out the noise, and brings back the handful of contracts with your name on them. They were yours all along.