How Tenderline Calculates Its Numbers

1. Three units, never blended

UK procurement has an administrative envelope (the procedure) and a commercial unit (the lot). Money and outcomes attach to awards and signed contracts. Tenderline never mixes these grains in one figure:

• Procedures (notices) — counted for pipeline volume, buyer and sector activity, routes and thresholds. A multi-region or multi-CPV notice counts in each region or category it names; hub pages say so and the counts are never summed across hubs.
• Lots — the unit of competition. Bid counts and award-vs-estimate pairs are always per lot; a seven-lot tender is seven contests.
• Award records / contracts — the unit of money. Values are VAT-exclusive (net) where published; VAT-only figures are counted, never summed, and never divided by 1.2. Framework and DPS ceilings and values of £1bn or more are excluded from sums as notional.

2. Competition figures

"Typical bids per lot" and "Single-bid lots" are computed on contested contract notices — direct awards (including PA 2023 transparency notices) report one bid by definition and framework / DPS lots draw a different population, so both stay out. The window is the last 36 months on sector, region and market pages, and 24 months on authority dossiers. On Procurement Act 2023 notices the bid count is the tenders assessed at the final stage, not every submission received. Award-vs-estimate pairs compare like with like (net against net, or gross against gross), need a published lot estimate, a single active award with one canonical supplier, and a gap under 60%.

3. Sign convention

Savings and price movement are (estimate − award) ÷ estimate. Positive means the award came in below the published estimate; negative means above it. Medians within ±0.5% are shown as "at the estimate" — many buyers publish the award at exactly the estimate.

4. Who wins here

Supplier tables on region, sector, market and platform pages are ranked by the number of distinct procedures a supplier won, not by award records: a supplier appointed to hundreds of call-offs under one framework should not outrank a supplier winning hundreds of separate competitions. Award records (every single-supplier award, frameworks and DPS panels included, all retained years) are shown alongside as a volume column, and each surface states its population.

5. Windows and samples

Every figure carries its own window and sample size: hub competition blocks use 36 months, authority dossiers 24 months, supplier award facts are all-time, and the retender radar counts signed contracts by their published end dates. Sample tiers follow the printed n: robust at 5+ observations, "moderate" at 3–4, "indicative" at 1–2; below the printed threshold a figure is muted, never invented. Medians are computed over the stated observations directly — never a median of medians — and when a more specific category has too few observations we fall back to a wider category and label it, rather than blending.

6. Matched shortlists and scores

"Who you are likely to meet" lists are model output, not published fact: suppliers on this authority's earlier similar contracts (matched by title, lot, buyer and CPV) and suppliers with a matching track record. "Match" and "relevance" are 0–100% heuristic scores, not probabilities. Find a Tender does not publish who is bidding; contact details, payment behaviour and recommended prices are never shown.

7. Sources and corrections

All facts derive from official Find a Tender notices published under the Open Government Licence v3.0, normalised nightly. Figures update as notices are amended; award values follow the award's own published value, else its contract's. If a figure looks wrong, the notice it came from is one click away — and tell us and we will recheck it.