RFP Evaluation Criteria That Survive Protest

This is Part 5 of The Solicitation Author’s Handbook, a vendor-neutral series for people in offices of government who write solicitations. Part 4 covered requirements vendors can answer. This part covers how you will judge the answers, which must be decided before you publish, not after you read. Educational information, not legal advice.

The iron rule: criteria before release

Every protest-resistant evaluation shares one property: the criteria, their weights, and the scoring method were fixed and published before responses arrived, and nothing about them moved afterward. The moment criteria shift after opening, even innocently, even to “reflect what we learned,” the award becomes an argument, and the record reads exactly like the thing procurement law exists to prevent. Decide before release. If reading proposals genuinely teaches you the criteria were wrong, the honest remedies are amendment before the deadline or cancellation and reissue, never quiet reweighting.

Publishing the criteria and weights in the solicitation itself costs nothing and buys two things: vendors write to what you value, and your file shows the rules preceded the players.

Weights are a policy statement

The split between technical merit and price is not arithmetic, it is a declaration of what the purchase is for. A 70/30 technical-to-price weighting says quality differences matter more than price differences within the competitive range; 30/70 says the opposite. Neither is wrong, but each fits different purchases, and the weight you choose should be a deliberate answer to one question: over the life of this contract, which failure is more expensive, paying somewhat more, or getting somewhat less? Score price with a disclosed formula (lowest price earns full points and others earn proportionally is the common one) so no one can suspect the price score of judgment.

Best value is the name most codes give this balancing; lowest responsive bid remains right where Part 2 put it, purchases where the specification really does capture everything that matters.

Rubrics: what a 3 means

The quiet failure mode of evaluation committees is five honest people scoring on five private scales. The repair is a rubric that defines each score level in words before anyone reads a proposal: a 5 demonstrates the outcome with evidence and named experience; a 3 addresses the requirement credibly but generically; a 1 restates the requirement back as an assertion. Write the rubric per criterion, not globally, and pin scores to evidence in the proposal (“page and section”) rather than impressions. Rubrics do not remove judgment; they make judgment legible, alike across evaluators, and defensible eighteen months later when someone who was not in the room asks why.

The committee

Small and odd-numbered works: three to five evaluators with genuinely different vantage points, program, technical, and financial at minimum. Before anyone receives proposals, each member signs a conflict and confidentiality screen, and anyone with a disqualifying relationship steps out, visibly. Score independently first; discuss second; document any score a member changes after discussion with a sentence about why. Consensus scoring has its place in some codes, but independent-then-reconcile leaves the record protests respect most: individual sheets, a reconciliation memo, and arithmetic anyone can repeat.

Two roles sit outside the scoring: a facilitator who owns process and the file, and, where your office provides it, procurement counsel on call for questions that start with “are we allowed to.”

Demonstrations, references, and the record

If scripted demonstrations or oral presentations are part of evaluation, they are scored extensions of the written process: same script for every vendor, same time box, rubric written in advance, notes filed. Reference calls likewise: same questions, asked of every finalist’s references, with answers recorded. The discipline is not ceremony. Every one of these artifacts is a page of the story you will tell if the award is questioned, and the story must have been written as it happened, not reconstructed afterward.

Debriefs are trust infrastructure

Offer losing vendors a debrief, and treat it as the beginning of the next procurement rather than the end of this one. A good debrief walks the vendor through their own scores against the rubric, factually and without comparison to the winner beyond what your code permits. Vendors who understand why they lost respond again next time, sharper; vendors who are ghosted respond with records requests, protests, or silence, and the silent ones were often the ones you wanted back. The evaluation file you built above makes debriefs cheap; ten minutes of preparation per vendor, because the reasons are already written down.

What the award memo contains

One document closes the process: the recommendation memo. It states the criteria and weights as published, the scores as reconciled, the price analysis, any responsibility determinations, and the recommendation, in language a newspaper could quote without embarrassing anyone. If you cannot write that memo comfortably, the evaluation is not finished, whatever the spreadsheet says.

The evaluation checklist

• Criteria, weights, and scoring method were published in the solicitation and never moved.
• Price scoring uses a disclosed formula.
• Per-criterion rubrics define score levels in words, and scores cite proposal evidence.
• Evaluators scored independently before discussing, with conflicts screened in writing.
• Demonstrations and reference checks followed identical scripts per vendor.
• The award memo could be read aloud at a public meeting.

Part 6 covers the clock: building the solicitation calendar, running the question-and-answer period without amending scope by accident, and getting cleanly from close to award.

The Solicitation Author’s Handbook is published by GovSoft as a public resource for offices of government. It is educational information, not legal advice.