Certvas
Signal · SOE-Procurement v1

What the state actually buys.

Government tender flow resolved to the canonical buying entity — how much each state or parastatal body puts to market, how many distinct suppliers win its work, and how concentrated that supplier base is. Every row carries its resolution confidence and a point-in-time window.

Grain: buyer entity (dataco_id) · point-in-time (release_date) · AF-TENDER (OCDS) × the entity spine · also via cv.signals.get("certvas-soe-procurement-v1") and /v1/signals/certvas-soe-procurement-v1 · 4 provenance guarantees

Why buyer-side

This is the direction the data joins.

Measured on a real Zambian window (21,674 tenders, 1,859 parties): 3 of 652 buyers resolved to the entity spine — and 0 of 1,207 suppliers. That asymmetry is structural, not a coverage gap. The organisations holding an LEI in these markets are state and parastatal buyers — central banks, mining holdings, utilities — while the suppliers winning their tenders are private SMEs with none. So this signal reads the side the spine can actually resolve.

FieldTypeDescription
dataco_id / entity_namestringCanonical buying entity and its legal name.
jurisdictionstringEntity jurisdiction (matching is within-country only).
tender_countintegerDistinct tenders published by the entity.
total_tender_value / currencynumber / stringNominal advertised value; mixed-currency and sparse in OCDS.
award_count / supplier_countintegerAwards ingested against those tenders, and distinct winning suppliers.
top_supplier_share_pctnumberShare of awards (by count) won by the single largest supplier.
first_tender_date / last_tender_datedatePublication window bounds.
match_tier / match_score / name_variantsstring / number / integerResolution confidence, and how many published spellings collapsed into this entity.
observed_atdateWhen the tender became knowable in Certvas gold.

Honest coverage: only buyers whose published name resolves to a canonical entity appear, so this is not a census of state procurement — most municipalities and smaller agencies are absent and national totals are understated. Awards lag tenders, so an entity can show real tender flow with award_count = 0; top_supplier_share_pct is then null rather than guessed. Concentration is computed on award counts over awards Certvas has ingested, so a high share can reflect thin award coverage rather than genuine concentration. It is a screening prompt, not a finding — a single-supplier entity may simply be procuring a specialised good, and nothing here asserts wrongdoing by any named organisation.