Data & methodology

Award amounts, awarding agencies, and competition fields (offer counts, set-asides, extent competed) come from the public USAspending.gov API.

Notice text, notice type, and deadlines come from SAM.gov.

Both are official US government databases. Hashike is not affiliated with or endorsed by the US Government, Department of Defense, or SAM.gov.

Dollar figures are the obligated amount — what has actually been committed under the contract — not the potential/ceiling value, which is typically larger.

Figures like “total awarded” sum the obligated amounts of every award matching the defined scope (NAICS/PSC, time window, place of performance) — not the size of any one opportunity.

Recipient location is where the contracting company is registered; place of performance is where the work actually happens. A US-based company can win a Japan-performance contract, so these are not the same thing — Hashike's “Japan-based firm” figures use recipient location.

The market analysis on an opportunity page covers the trailing 48 months, the same NAICS code (falling back to PSC), Japan place-of-performance, and prime contract awards. The exact number of awards matching those filters is read from the count API and shown. Where that number fits within the analysis limit we analyze every one of them; where it does not, we analyze every award at or above a stated dollar floor; and only if that is still too large is the set a sample, which the page says explicitly.

Offer counts, set-asides and contracting offices exist only on per-award detail records, which we retrieve for a subset of the analyzed awards. Every percentage derived from them is calculated over the records that actually carry the field — never over the wider set — and each figure is shown next to the population and record count it was computed from.

A company's award-history page examines up to its 100 most recent awards. For companies with more history than that, the total and first-award date shown are a floor, not the true figures — the page flags this when it applies.

Supplier and buyer “concentration” uses the Herfindahl-Hirschman Index (HHI), the standard economics measure: each participant's percent share, squared, summed — a number from 0 to 10,000. Per the bands used by US antitrust agencies (DOJ/FTC), below 1,500 is read as unconcentrated, 1,500–2,500 as moderately concentrated, and above 2,500 as highly concentrated.

On an opportunity page, concentration is computed by attributing every analyzed award to a resolved supplier, so the figure is exact for that population and the shares sum to 100%. Share of observed value and share of award count are different metrics that can diverge sharply for the same supplier, so both are shown.

On company pages, concentration comes from USAspending's category API, which returns only the largest names by dollar volume. The remainder is folded in as a single “other” block, which overstates true concentration when that remainder is many small entities — so that figure is a ceiling, not a point estimate.

Concentration describes market structure. It was not built as, and should not be read as, guidance on whether to bid.

The market analysis on an opportunity page does resolve name variants. Records sharing an official identifier (UEI) are merged as one entity; without a UEI, records are merged only on an exact match of the company name after removing case, punctuation and legal-form suffixes (Ltd, Co., Ltd, Inc and so on). Similar names are never merged — “NIPPON STEEL” and “NIPPON STEEL ENGINEERING” stay separate — because inventing a market structure that does not exist is worse than leaving a company split. The number of entities resolved by name alone is reported in each page's data-quality panel.

Company award-history pages are a separate path: they match on USAspending's recipient-name text and do not resolve variants, so a company whose awards are recorded under more than one spelling can be undercounted there.

The opportunity feed refreshes daily. The per-code market analysis behind each opportunity page is recomputed weekly, and a newly seen code is analyzed in the next daily run; every page shows when its data was last refreshed. Company award histories are fetched live from USAspending each time the page loads.