There are 575,165 public contracts due to expire in the next twelve months, each with a named incumbent supplier. The number the average procurement team is tracking outside its own contract book: close to zero. What sourcing teams miss — and why the gap is structural, not lazy.
Every procurement leader can tell you, to the dollar, what their organisation is spending right now. Which suppliers are on contract. What renews this quarter. Which categories are over budget. The internal book is the one thing procurement tooling has been built to see clearly for thirty years.
Ask the same leader a different question — which of your competitors' contracts come up for renewal in the next twelve months, and who holds them today — and the answer is silence. Not because the information is secret. Because nobody assembled it.
We maintain 82,887,032 public contract awards drawn from 124 government and multilateral sources worldwide — USASpending and SAM.gov, the EU's TED system, the UK's Contracts Finder, Brazil's PNCP, Mexico's Compranet, Ukraine's Prozorro, Australia's AusTender, and hundreds of national and sub-national portals besides — normalised into a single engine. This week we ran a query nobody outside a procurement function ever thinks to run: not who won, but who is about to lose.
Procurement technology optimises the spend you already own. Contract lifecycle management tracks your contracts. Spend analytics categorise your invoices. Supplier management scores your vendors. Every one of these tools is a mirror: it shows the organisation itself, in high resolution, and nothing beyond the glass.
That is a reasonable design for running a procurement function. It is a terrible design for growing one. Because the single most predictable event in all of procurement — a contract reaching its end date — is invisible the moment it belongs to someone else. Your incumbent's other contracts, your competitor's book, the category leader whose three-year deal with a government buyer quietly lapses in March: all of it sits outside the mirror.
The result is that sourcing strategy runs on the two worst possible information sources — the RFP that has already been published (too late to shape) and the relationship the account manager chooses to mention (too partial to trust).
The blind spot is not a coverage gap. It is a framing error. Procurement data is treated as a record — an audit trail of what was bought — when the same data, turned ninety degrees, is a forecast of what must be bought again.
A contract award is not just a receipt. It carries an end date, a named supplier, a buyer, and a category. Aggregate enough of them and the end dates stop being administrative trivia and become a calendar: a dated, named, sortable schedule of every market about to reopen. The award record you file and forget is the exact signal your competitor would pay for — pointed at your incumbents.
"The contract you file and forget is the same contract your competitor would pay to see — pointed at your incumbents."
Across the dataset, 11,979,782 awards carry both a contract end date and a named supplier. Most are historical — contracts that ended years ago. The part that matters to a sourcing team is the forward wall: contracts with an end date still ahead of them.
There are 988,510 of those — just under a million public contracts, each with a dated future expiry and a named incumbent, spread across 73 of our 226 covered buyer countries and territories, and 180,296 distinct suppliers. This is the addressable market that no contract-management system contains, because it is composed almost entirely of contracts that belong to somebody else. It is heavily concentrated in the jurisdictions that publish most openly: the United States alone accounts for 643,902 of the 988,510 — roughly two in three — followed by the United Kingdom, Germany and Australia.
The near term is where it bites. Of that wall, 575,165 contracts expire within the next twelve months — the renewal cliff. This is a rolling twelve-month window, running to July 2027, which is why it covers the rest of this year plus part of 2027 rather than mapping cleanly to a single calendar-year row below. Each one is a market that is, on a known date, guaranteed to reopen, with a named holder who will fight to keep it and a field of challengers who mostly do not yet know it is coming.
| Expiry year | Contracts | Notes |
|---|---|---|
| 2026 (H2) | 335,164 | remainder of this year |
| 2027 | 473,495 | the peak of the visible wall |
| 2028 | 83,797 | |
| 2029 | 48,218 | |
| 2030 | 22,220 | |
| 2031 | 10,050 | |
| 2032–2041 | 15,566 | the long tail |
The wall thins as it extends, and that is a feature of contracts, not a hole in the data: most public contracts run one to three years, so the near years are dense and the far years naturally hold fewer live commitments. The rows sum to the full 988,510.
Two honest notes on the data. First, these are counts of contracts, not a sum of their value — deliberately. Reported contract values in raw public data are unreliable in aggregate: they mix true awards with framework ceilings and outright misparses, and a naive total is dominated by a handful of them (more on that below). Counting dated, named contracts is the measure that survives scrutiny. Second, contract end dates are entered by humans, and humans enter placeholders. The wall is bounded to contracts ending on or before 24 July 2041 — fifteen years out — so 4,839 records with end dates beyond that horizon, including the classic 9999-12-31, are excluded from every figure here. Fifteen years is where genuine incumbency lives. This is a floor, not a ceiling: it counts only contracts where a real end date was disclosed. The true renewal market is larger, not smaller.
None of this data is private. Every one of those 988,510 contracts was published by a government buyer, in the open, as a matter of law. The blind spot is not access. It is assembly.
The information is scattered across roughly 115 separate portals that do not agree on anything — not the language, not the date format, not the currency, not the spelling of the supplier's name. A German municipal award, a US federal IDV, a Brazilian pregão and a UK framework call-off describe the same underlying event — a buyer committing money to a named supplier until a known date — in incompatible schemas. Reassembling them into one calendar is unglamorous, expensive, and the entire game.
And the raw data actively lies to you if you let it. In the twelve-month cliff, the single largest contract by reported value is a $343 billion award to a small Greek supplier — a figure that is not real, but a parsing artefact from a source that stored a ceiling or a concatenated field as a contract value. The next-largest is a UK record whose "supplier" is the literal text "a list of successful suppliers can be found…".
A parsing artefact: a small Greek supplier credited with a $343 billion award because a source stored a ceiling or a concatenated field as a contract value. A dataset that hasn't done the unglamorous work will confidently tell you this is the biggest deal in the market. Ours flags and excludes it — which is exactly why we count contracts and not naive dollars.
Two disciplines make the difference between a dataset and a liability. The first is deterministic entity resolution: matching "Thales SA", "THALES", and "Thales Group" to one incumbent using exact and normalised identifiers, never similarity-guessing — because a match that turns Georgia into Serbia is worse than no match at all. The second is point-in-time integrity: an expiry date is only trustworthy if it came from a real dated disclosure, not a default. Where the source is silent, the honest answer is a null, not a fabricated date that puts a phantom contract on the calendar. Coverage you cannot trust is worse than coverage you do not have.
A named, dated renewal calendar changes three plays at once.
Defend. An incumbent who knows, twelve months out, exactly which of their own contracts sit on the cliff — and which competitors hold adjacent, simultaneously-expiring deals with the same buyer — is defending on offence. Renewal stops being a surprise and becomes a scheduled campaign.
Attack. A challenger can, for the first time, target the specific contracts a rival is about to have to re-win, in the categories and geographies they choose, ranked by expiry date. Business development stops being a cold-outreach lottery and becomes a dated pipeline.
De-risk. Turn the same calendar toward your own supply base and it reads as a risk register. Concentration shows up as clusters of critical contracts expiring in the same quarter with the same incumbent; single points of failure surface before they become incidents, not after. This is where procurement intelligence and third-party risk become the same dataset viewed from two directions.
Universe: Valan's master_awards dataset — 82,887,032 public contract awards ingested from 124 distinct sources (counted at feed level; roughly 115 parent portals once sub-feeds such as TED_SUBCONTRACTS and USASPENDING_SUBAWARDS are collapsed) and normalised into a single DuckDB engine, live as at 24 July 2026.
The forward wall counts awards with a named supplier and a contract end date parseable as a real date falling between today and 24 July 2041 (fifteen years out); 4,839 records with end dates beyond that horizon are excluded. Supplier identities are resolved deterministically, using exact and normalised matching only — no similarity or inferred matching.
Contract values are reported by counts, not summed dollars: in the twelve-month cliff the ten largest reported values account for roughly 41% of the nominal total and are dominated by ceiling amounts and parsing artefacts (including a spurious $343bn record), so a value aggregate would materially mislead. Figures will move as contracts are awarded, amended, and expire. All sources are open, official primary records.
This is structured data, not a strategy. The 575,165 contracts on the twelve-month cliff do not tell you which one to chase or which supplier to drop. They tell you where the markets are about to reopen, who holds them today, and when the clock runs out — current, global, and assembled before the RFP lands.
What you do with that depends on your category, your capacity, and your own judgement. What we provide is the raw material the internal mirror was never built to show: not the spend you already own, but the market about to come loose.
New research from Valan's 82.9M-award procurement dataset, delivered as it publishes. No noise.