Know the day a competitor wins.

Arcthane reads every federal award, recompete and solicitation the day it publishes. Capture resolves it to the actual company behind it, public or private, for capture and BD teams. Terminal reads the same record as an investor: signals, scores, intraday. Know first, decide yourself.

7,554 realized outcomes on the public record · re-scored every 2 hours · research, not advice

Archive coverage

198,639+Federal awards
25,000+Companies, public and private
60Months of contract coverage
2010FDA approvals to
1994M&A records to
17,714Scored events

USASpending, SAM.gov, SEC EDGAR, Congress.gov, FDA, sanctions lists, and 20+ sources in total, polled continuously and re-scored every 2 hours. Records are deduplicated against prior notices and resolved to the company itself, public or private, not just the contract line, before scoring. The depth is per source, not only in aggregate: FDA approvals back to 2010, M&A history to 1994, and 60 months of contract-level detail, all cross-referenced against the same company graph.

One signal, five fields, every time.

Timestamp, source, headline, strength, sizing. The order never changes, so the muscle memory transfers between the brief and the Terminal.

ConfirmedLMTLockheed Martin86

$7.8B sustainment award, 6.4% of market cap. Scored within hours of the award posting.

Timing same-dayMateriality transformativeAward / mcap 6.4%Action buy watch

Fig. 01 · a signal, as your feed renders it · illustrative example, not a live signal or a performance claim

Three moves, repeated every two hours.

The same engine feeds both products: Capture renders the record itself, by company, for capture teams; Terminal renders the scored read for investors.

01 · Ingest

The entire public exhaust of federal money: awards, solicitations, subcontracts, budget lines, filings, approvals. Millions of documents. No manual curation.

02 · Score

Materiality against market cap and revenue. History of what similar awards did to similar stocks. The macro regime it lands in. One number, 0-100.

03 · Time

Same-day reads get flagged fresh. Late reads get flagged honestly. A signal that says "likely priced in" is the one that saves you money.

The Terminal, as it renders

app.arcthane.com/feed
Arcthane Terminal, the unified signal feed

Fig. 02 · the Terminal feed, as it renders live

Five ways the future leaks.

Most tools show you the news. The news is the last step of a chain that starts months earlier: in budgets, expiring contracts, and filings. Arcthane labels every signal by where in the chain it was caught.

Most of it is noise. We publish exactly how much.

Across 1,219 realized contract outcomes, the average signal underperforms its sector (-2.9% median over 60 trading days). The contract-award universe is a trap, and the product is the filter that says which slice isn't. Out of sample in 2025-26, top-ranked signals beat their sector 53% of the time against a 47% base rate, median +3.1%. That edge is attributable to award size, with wide dispersion (worst decile -25.7%).

We publish the losing rows, the baseline, and the out-of-sample artifacts. A track record you can't audit is marketing; this one updates from live data every hour, and the honesty is the moat.

Open the full record
Out-of-sample ranker comparison, fit through 2024 scored on 2025-26: frozen linear model, live ML score, and award size. Hit rate against base rate with n.
Ranker · OOS 2025-26Hit (60d)BaseMedian αn
Frozen linear model40%47%-8.6%284
Live ML score46%47%-5.8%284
Award size (shipped)53%47%+3.1%284

fit ≤2024 → scored 2025-26 · α = return minus sector ETF · worst decile -25.7% · full protocol at /methodology

The record is public. The reading isn't.

Three minutes to set up. Billed monthly, cancel any time, self-serve.

Get started