Your evaluation model is the contract you will live with for ten years
By the time bids land, most of the award decision has already been made. It was made when someone chose the weightings, wrote the questions and set the scoring bands. The May 2021 Bid Evaluation guidance note understands this better than most procurement teams do: it spends more pages on designing and testing the model than on scoring itself. Read carefully, it is a warning that evaluation failures are design failures, and they are baked in before the contract notice is ever published.
The maths quietly overrides your intentions
Teams announce that quality matters most, publish a 70/30 weighting, and then design quality questions so generic that every competent bidder scores the same. The guidance names this trap directly: if quality criteria do not differentiate, the procurement is decided on price regardless of the published weighting. Bid bunching is not bad luck, it is a design outcome, and the note suggests concrete countermeasures like non-linear scoring thresholds and even numbers of bands.
Relative price scoring gets similar treatment. It can hand the lowest bidder a decisive advantage and distort scores when there are few bidders, which is why the guidance says it should not be used without a specific business reason approved by the commercial lead and the SRO. Alternatives like price per quality point evaluate each bid against an absolute standard rather than against whoever happened to bid lowest.
Cheap bids are the expensive ones
The most sobering passage in the note lists what happens after a too-low bid wins: suppliers cutting costs instead of delivering quality, demanding contract changes, managing to the letter of the contract during a crisis, and departments injecting emergency funding to protect critical services. Some of those rescues can even breach the regulations as substantial modifications. The bias towards low cost bids is not thrift, it is deferred spending with interest.
The defence is knowing what the service should cost before anyone bids. That is why the Should Cost Model is mandatory for complex outsourcing, and why any bid more than 10 per cent below the average or the model must be referred for central scrutiny. Abnormally low bids trigger a legal duty too: ask the bidder to explain, and reject only if the explanation fails. None of this works if the model is built after the bids arrive.
Moderation is where challenges are won and lost
Unsuccessful bidders do not usually challenge the outcome, they challenge the process. The guidance reads like a checklist compiled from lost court cases: the same evaluator scoring all responses to a question, clarifications only in writing through the commercial team, consensus scores that are reasoned rather than averaged, and a written justification for every score change in moderation.
The moderator's red flag list deserves printing out: comments about past experience with a bidder, comparisons between bidders, pasted scoring descriptors, justifications that only make sense with the evaluator in the room. Each one is a defect that surfaces months later in a debrief letter or a courtroom. Catching them in moderation costs an afternoon. Catching them in litigation costs the procurement.
Treat the audit trail as a live deliverable
The note repeats one idea in almost every section: write it down, at the time, in language consistent with the criteria. Training records, conflict of interest forms, individual scores, moderation minutes, clarification logs, the final evaluation report under Regulation 84. Teams that assemble this as they go finish award-ready. Teams that leave it to the end discover gaps precisely when a standstill letter has started the clock.
The practical move is to run evaluation inside a system that records the process as a by-product of doing it, rather than treating documentation as a separate job. However you achieve it, the test is simple: could you hand your records to a sceptical outsider and let them reconstruct why the winner won? If not, the evaluation is not finished, whatever the scores say.
The takeaways
- Award outcomes are mostly determined by model design, so test the model with dummy bids before publishing it.
- Published weightings mean nothing if the questions and scoring bands cannot differentiate bids.
- Build the Should Cost Model before bids arrive and refer any bid 10 per cent below it or the average for scrutiny.
- Run clarifications in writing through the commercial team, never evaluator to bidder.
- Keep contemporaneous records that would satisfy a sceptical outsider; that is the real definition of a finished evaluation.
Want the full breakdown?
The complete explainer covers the key facts, the requirements in detail and a practical action list, free and printable in the Procurement Library.