What the evaluator will never tell you — anatomy of the scoring process

This article follows How to write a technical proposal that wins contracts, where we described encoding the signal to maximize the score, and Cognitive biases and tenders, where we showed that the bid manager is his own worst enemy. Here, we switch sides. We cross to the other side of the table — into the mind of the person scoring you.

The blind spot

Pick up any bid management guide. Any article. Any consultant. The topic is always the same: how to write, how to structure, how to win. The perspective is systematically that of the bidder.

Nobody talks about the other side.

How they read you. How they score you. How they compare you. How the panel works. What biases operate in the mind of the person awarding 14/20 instead of 16/20. Why the same technical proposal gets different scores depending on whether it is read first or fifth. Why a technically superior submission sometimes ends up behind a weaker one.

This is a massive blind spot. And it is structural.

Public procurement evaluators do not write guides. They do not publish lessons learned. They have neither the time, the incentive, nor sometimes the right to do so. The scoring process is a black box — and bidders optimize their responses for a box they have never opened.

It is like preparing for an exam without ever seeing an answer key. You know the syllabus. You study. But you do not know how the examiner reads your paper, in what order he scores it, what irritates him, what impresses him, what he looks for when he cannot find it. You are optimizing in the dark.

Key takeaway: The entire existing body of work talks about "how to write a good response." Nobody talks about "how that response is read, scored, and compared." Until you understand the scoring process, you are optimizing your technical proposal for a target you have never seen.

---

How the evaluator reads your bid — the real process

Forget the image of the conscientious reader going through your proposal from page 1 to page 200. That never happens. The evaluator has five submissions to score. Sometimes seven. Each one runs between 80 and 300 pages. He has two to three days. He has a printed grid, a pen, and mounting fatigue.

Here is what actually happens.

Pass 1 — The sort

The evaluator opens the file. He looks at the table of contents. He checks that the document is structured, paginated, compliant with the required format. He reads the executive summary — if there is one. He skims the first pages of each section to get an overall sense.

What he is looking for: "What is this about? Is this serious? Does this candidate understand the contract?"

This is not reading. It is scanning. It takes between 10 and 20 minutes for a 150-page submission. And by the end of this pass, the evaluator already has an impression. Not a score. An impression — positive, neutral, or negative. That impression will color the entire rest of the reading.

This is the anchoring bias in action: the first impression weighs more than subsequent information. A generic executive summary — "Our multidisciplinary team is committed to supporting your transformation" — anchors the evaluator in the negative. A precise executive summary — one that names the specific issues of the contract, the identified risks, the response strategy — anchors in the positive. The rest of the submission is read through that filter.

Pass 2 — The scoring

This is where everything is decided.

The evaluator takes his scoring grid. Criterion by criterion, he searches the proposal for the section that addresses it. He scans headings, subheadings, the first sentences of each paragraph, tables, call-out boxes. He is looking for addressable response elements — a piece of information he can map to a line on his grid and assign a score to.

If he finds the information easily, he scores it. If he does not find it within 30 seconds, two things happen. Either he searches a bit longer — if the submission made a good impression in pass 1. Or he marks "insufficient" or "not addressed" and moves to the next criterion — if the impression was neutral or negative.

Thirty seconds. That is the time you have for each answer to each criterion to be found and understood. Not because the evaluator is lazy. Because he has 1,500 pages to process in two days and scanning is a channel capacity constraint, not a choice.

Pass 3 — The deep review

This one is rare. It only concerns the 2-3 shortlisted submissions — those in the running for first place. The evaluator goes back to the key sections, verifies figures, compares approaches across submissions. He is looking for differentiators.

Most submissions will never see this third pass. They are eliminated — or definitively ranked — at pass 2.

Key takeaway: The evaluator does not read your bid. He scans it in three passes. The first anchors an impression. The second produces the score at 30 seconds per criterion. The third — which most submissions never see — separates the finalists. If your proposal is not built to survive a 30-second scan per criterion, it is built to lose.

---

The weighted scoring grid — what nobody explains

Every serious bidder knows the evaluation criteria and their weightings. They are public — public procurement law requires it (Article R2152-7 of the French Code de la commande publique). "Technical value: 60%. Price: 40%." Or: "Methodology: 40 pts. Human resources: 30 pts. Understanding of needs: 30 pts."

What nobody says is that these criteria are only the visible layer. Underneath, there are the sub-criteria — and that is where the score is actually produced.

Internal granularity

The tender notice states "Methodology — 40 points." But how are those 40 points distributed? The contracting authority has a choice. It can break them into:

• 8 sub-criteria of 5 points (fine granularity)
• 4 sub-criteria of 10 points (medium granularity)
• 2 sub-criteria of 20 points (coarse granularity)
• No sub-criteria — 40 points awarded in a single block (holistic evaluation)

The decomposition changes everything.

With 8 sub-criteria of 5 points, an average submission can accumulate steady 3/5 and 4/5 scores and finish at 28/40. An excellent submission on some aspects but weak on others can end at the same score — peaks and valleys cancel out.

With 2 sub-criteria of 20 points, variance explodes. An evaluator who gives 18/20 on the first sub-criterion and 10/20 on the second produces the same total of 28/40 — but the score reflects a radically different profile.

With a single-block 40-point score, subjectivity is maximal. The evaluator reads the entire section and assigns an overall impression. No fine-grained traceability. No way to tell which elements counted.

The problem for the bidder

Sub-criteria are not always published. The contracting authority is not required to do so (unless they constitute autonomous criteria in their own right — per case law, Conseil d'Etat, 2016). Result: you know the criteria but not the detailed scoring grid.

The bid manager who does not know how many sub-criteria make up "Methodology — 40 pts" has two options. Either he covers everything uniformly — this is the guaranteed 12/20 strategy. Or he identifies the likely sub-criteria by analyzing the detailed requirements of the specification and structures his proposal as a mirror — this is the proposal-aligned-to-the-scoring-grid strategy that targets 17/20.

The specification gives the clues. If the methodology section of the requirements document details eight distinct themes (governance, qualification process, incident management, change management, reporting, document management, quality plan, knowledge transfer), it is reasonable to anticipate that the internal grid reflects those eight themes.

Key takeaway: The bid manager who does not structure his proposal as a mirror of the scoring grid forces the evaluator to search. And an evaluator who searches is an evaluator who penalizes — not out of bad faith, but out of time pressure. The weighted scoring grid is not an administrative detail. It is the invisible architecture that determines your score.

---

The evaluator's biases

The bid manager has his biases — we have done the detailed anatomy. But the evaluator has his too. And his have a direct impact on your score.

The fundamental difference: the bid manager can work on his biases. He can formalize his process, use checklists, rely on a cognitive system. The evaluator operates within a constrained framework — limited time, multiple submissions, institutional pressure — where biases are not only present but structurally unavoidable.

Anchoring bias — the first submission calibrates the scale

The evaluator reads the submissions in a given order. The first one he opens becomes, unconsciously, his reference point. If it is the best submission in the lot, all subsequent ones suffer by comparison. If it is the weakest, the others benefit from a favorable contrast effect.

The evaluator does not choose the reading order. He picks up the pile. And the first impression of the first submission calibrates the scoring scale he will apply to all the others. This is Tversky and Kahneman's anchoring bias applied in real-world conditions — and no scoring grid, however detailed, eliminates it.

Halo effect — one brilliant section colors the rest

The evaluator who reads an exceptional "Understanding of needs" section — precise, factual, naming the risks, reformulating the stakes — develops a positive impression of the entire submission. When he arrives at the "Human resources" section, which is adequate but unremarkable, he tends to overrate it. "This candidate is serious, their team section must be solid too."

The reverse is true. A weak section — generic, recycled, without evidence — contaminates the evaluation of subsequent sections. The evaluator reads the rest with a negative prior. Even a good methodology section will be read with more skepticism after a botched understanding of needs.

The halo effect works in both directions. And it operates most powerfully on the first sections read — which explains why the executive summary and the understanding of needs are the most strategic sections of the proposal. Not because they weigh the most in the scoring grid. Because they calibrate the reading of everything else.

Cognitive fatigue — the 5th submission does not get the same chance

An evaluator who starts his day is attentive. He reads the details. He notes the nuances. He distinguishes a 14 from a 15. Four hours and three submissions later, his discrimination capacity has dropped. Studies in cognitive psychology (Baumeister, R.F. et al., Ego Depletion, Journal of Personality and Social Psychology, 1998) show that decision quality degrades significantly after a series of demanding tasks.

The fifth submission is read with 60% of the attention given to the first. Subtleties are missed. Fine arguments are skimmed. Long sections are scanned even faster. The score reflects not only the quality of the submission, but also when it was read in the pile.

The bidder cannot control this — he does not choose the reading order. But he can adapt to it: a submission built to be understood in degraded mode — with opening sentences that carry the message, tables that concentrate the evidence, headings that summarize the section — withstands cognitive fatigue better than a submission that requires careful reading to be appreciated.

Conformity bias — the rapporteur influences the panel

In an evaluation panel, members are not independent. The rapporteur — the one who read the submission in depth and presents his analysis — influences the co-evaluators. This is Asch's conformity bias: in group settings, individuals tend to align their judgment with whoever speaks first.

If the rapporteur is enthusiastic, co-evaluators will tend to confirm. If he is critical, they will look for flaws rather than strengths. The final score reflects less the collective opinion than the rapporteur's opinion amplified by conformity.

Key takeaway: The evaluator is not a machine. He is subject to the same cognitive biases as the bid manager — anchoring, halo, fatigue, conformity. The difference: the bidder cannot correct the evaluator's biases, but he can build a submission that resists them. A submission that anchors positively, that gives no foothold to negative halo, that survives fatigue, and that is easy to defend in panel.

---

Panel dynamics

The evaluation panel is the moment where your score crystallizes. It is also the least understood moment of the entire process.

The rapporteur presents — the others have skimmed

In most panels, each lot or criterion has a rapporteur. He is the one who actually read the submissions in detail. The other panel members? They skimmed. Sometimes read the executive summary. Sometimes nothing at all. They arrive with a blank grid and listen to the rapporteur.

This is not negligence. It is a time constraint. A contract with five lots and five bidders is 25 evaluations to conduct. If every evaluator had to read every submission in depth on every criterion, it would take three weeks. They have three days.

Result: the rapporteur carries disproportionate weight. His analysis, his phrasing, his enthusiasm or restraint — all of it structures the discussion. Co-evaluators react to his presentation, not to the submission directly.

The debates focus on borderline cases

What does not get debated in panel: the excellent submission and the poor one. The 17/20 and the 9/20 are rarely contested. Everyone can see.

What gets debated: submissions between 12 and 15. The ones that are "adequate but..." The ones with strong sections and weak sections. The ones where the evaluator hesitates between two scores.

The panel debate follows a predictable pattern. The rapporteur presents. A member asks a question — "What did they plan for knowledge transfer?" The rapporteur searches his notes. If he does not find it immediately, the answer is "it is not clear in the submission." Not "it is not in the submission" — "it is not clear." The distinction matters: it means the information may be there, but it was not found. And in panel, information not found = information absent.

The soft consensus

A submission that triggers a debate is a submission in danger. Not because the debate is negative — but because the resolution mechanism is the soft consensus.

The rapporteur says "15." A member says "13." Another says "14." Nobody wants to be the outlier. Nobody wants to block the panel. A middle ground is found — 14. The middle ground is almost always lower than the rapporteur's score, because objections pull downward and positive arguments do not pull upward with equal force.

This is the fundamental asymmetry of the panel: it is easier to justify a reduction ("element X is missing") than to justify an increase ("the methodology section is truly exceptional" — "yes, but the transfer plan is missing"). The deficiency is concrete. The quality is subjective.

The best score is not the one that impresses a single evaluator. It is the one that is indisputable — that gives no foothold for an objection, no angle of attack for a panel member who wants to lower it. Every gap, every ambiguity, every unsupported claim is an open door to the soft consensus.

Key takeaway: In panel, a submission that triggers debate is a submission that loses points. The soft consensus pulls downward. The objective is not to impress — it is to be indisputable. Every section without a gap, every criterion covered, every piece of evidence provided is a point that nobody can take away.

---

What the evaluator cannot score — but that influences

There is a category of elements that appears in no scoring grid, corresponds to no sub-criterion, has no dedicated points — and yet influences the entirety of the score.

Layout and readability

The evaluator does not score layout. But he sees it. A well-spaced document, with clear headings, readable tables, clean pagination, adequate margins — is a document that says "we took this contract seriously." A compressed document, in 9-point font, with tables that overflow, inconsistent headings, and typos — is a document that says "we checked the box."

Layout has no dedicated points. But it acts as a silent multiplier of the halo effect. It colors the impression of pass 1. It facilitates or hinders pass 2. It influences the score of every criterion without ever appearing in the grid.

Internal consistency

The evaluator spots contradictions. Not always consciously — but the signal is powerful. If your "Understanding of needs" section identifies a risk of understaffing and your "Human resources" section proposes a minimal headcount without mentioning that risk, the evaluator registers the inconsistency. He does not score it explicitly. But the trust in the submission drops a notch.

Internal consistency is the invisible test: was this submission written by a single intelligence that commands the whole, or by three people who did not communicate? Recycled proposals are the most vulnerable — with each section coming from a different bid, contradictions are inevitable.

The feeling of understanding

The most powerful and least formalizable factor. When the evaluator reads a submission and thinks "these people understand my problem," the score goes up. Not by 2 points. By 3 to 5 points. Across all criteria. Because the feeling of understanding is not a criterion — it is a filter that amplifies or attenuates all criteria.

This feeling arises from accumulated details: a risk named that nobody else names, a reformulation that shows you read between the lines of the specification, a question asked during Q&A that hit the mark, vocabulary that mirrors the client's language rather than consultant jargon.

It is not in the grid. It is not measurable. But it is the factor that separates 14/20 from 17/20 when the substance is comparable. The evaluator cannot justify "I gave 17 because I felt they understood." But he can justify — and he does — "the understanding of context is thorough, the specific risks are identified, the methodology is tailored to the need." These are the measurable symptoms of a feeling that is not.

Key takeaway: Off-grid elements — readability, consistency, the feeling of understanding — have no dedicated points but influence every score. They are silent multipliers. A submission that masters them turns a 13/20 on substance into a 16/20 on the scoreboard. A submission that neglects them turns a 16/20 on substance into a 13/20 on the scoreboard.

---

Key takeaways

The scoring process is not the mirror of the writing process. Understanding how your bid is read, scored, and compared fundamentally changes how you build it.

Five lessons from the other side of the table:

1. The evaluator scans, he does not read. Three passes. The first anchors an impression. The second produces the score at 30 seconds per criterion. The third — which most submissions never see — separates the finalists. Every piece of information must survive the scan.

2. The weighted grid is an iceberg. The criteria are visible. The sub-criteria — which actually determine the score — are often invisible. Structuring your proposal as a mirror of the likely scoring grid is the single most valuable architectural decision of the entire bid.

3. The evaluator's biases are structural. Anchoring, halo, fatigue, conformity — they cannot be corrected, only circumvented. A submission built to resist them scores 3 to 5 points higher than one of equal quality that ignores them.

4. The panel pulls downward. The soft consensus penalizes submissions that invite debate. The objective is not to impress a single evaluator — it is to be indisputable for the entire panel.

5. The invisible factors are the most powerful. Readability, consistency, the feeling of understanding — no dedicated points, but a multiplier effect across all scores.

---

TenderGraph builds the technical proposal as a mirror of the scoring process — not the writing process. Every section is aligned with the likely scoring grid. Every criterion has an addressable response in under 30 seconds. Every claim is backed by traceable evidence. The system does not optimize for the bid manager who writes — it optimizes for the evaluator who scores. That is the only perspective that matters.

---

Read also:

• How to write a technical proposal that wins contracts — Encoding the signal section by section: structure by the scoring grid, opening sentences, evidence tables, strategic redundancy.
• Cognitive biases and tenders: the bid manager's worst enemy — The same cognitive mechanisms operate on both sides of the table. The bid manager and the evaluator are subject to the same biases — but in opposite directions.
• The myth of the executive summary — The exec summary is the anchor point of pass 1. If it is generic, the rest is read with a negative prior.
• What the specification does not say — The evaluator instantly recognizes the candidate who read between the lines. The feeling of understanding is born from this explicit inference work.
• Why your client references convince nobody — Logos do not score points. Contextual evidence does. The evaluator in pass 2 looks for facts, not names.
• The information revolution: signal, noise, and Shannon — The theoretical framework: entropy measures disorder, and fragmentation is an entropy machine. Shannon explains why the evaluator scans instead of reading.
• Responding to a tender: the method — The Go/No-Go is the first decision that impacts the final score: a bid you should not have made is a bid that will be read in pass 1 and eliminated.
• Your bid reviews are useless — The review on the eve of the deadline reproduces internally the conditions of the panel: fatigue, soft consensus, untraced decisions.
• Pre-sales is a command exercise — Understanding the scoring process is intelligence work on the opposing terrain. Without intelligence, no battle plan.
• Pre-sales skills in the age of AI — Putting yourself in the evaluator's shoes is the rarest and most profitable skill of the augmented bid manager.
• The oral defense: the moment where everything is decided — The oral defense is the extension of the scoring process — the moment where the evaluator verifies in person what he read in the submission.
• One tool for ten — A proposal structured as a mirror of the scoring grid requires a system that traces requirements end to end — not twelve disconnected tools.