How to Write a Winning Tender: What Evaluators Actually Score

Nobody reads your tender. It gets scored. If you want to know how to write a winning tender, that sentence is the whole syllabus. Three to five evaluators, a published marking scheme, a moderation meeting and a stack of rival bids to get through: every sentence you write either helps someone justify a mark or it is decoration.

Most losing bids are not badly written. They are competently written answers to questions nobody asked, padded with adjectives no evaluator can defend. The firms that win consistently are not the best stylists. They treat the evaluation matrix as the product specification and the word limit as a budget, and they put scoreable facts where their rivals put prose.

This guide covers what evaluators actually score, what separates a 3 out of 5 from a 5, and the three disciplines that put marks on the page: question fidelity, evidence density and word allocation. It assumes you picked the right contest in the first place. If you have not made that call properly, start with bid or no bid, because no amount of writing skill rescues a bad selection. And if the market itself is new to you, the complete supplier guide sets the scene.

How tender evaluation criteria actually work

Every formal tender publishes its evaluation criteria and weightings up front, and almost every losing bidder skims past them to get to the questions. Slow down there. The weightings are the buyer telling you, in numbers, exactly what they care about.

A typical split is 60% quality and 40% price, or 70/30 in quality led markets, with social value carved out as its own slice: at least 10% on central government tenders under the government's procurement policy notes. Under the Procurement Act 2023, buyers now award to the "most advantageous tender". The word "economically" was deliberately dropped from the old formula, and the message to evaluators is that quality and social value are allowed to carry the day.

Quality questions are scored on a ladder, usually 0 to 5 against published descriptors. The wording varies between buyers. The shape never does:

• 5: fully answers the question, specific method, strong evidence, no reservations
• 3: answers the question, but generically, with thin evidence and minor concerns
• 1: partially relevant, mostly assertion
• 0: does not address the question

Then comes moderation, the part most bidders never think about. Evaluators score independently, then meet to agree a consensus mark, and every score has to be defended with reference to your text. A moderated score must survive on the page alone. If the evidence is not on the page, it does not exist. Your reputation, your website, the warm meeting you had with the service manager in March: none of it is admissible.

The gap between 3 and 5 decides contests. A 3 says "they can probably do this". A 5 says "here is exactly how they will do it, where they have done it before, and how they measured the result". On a question worth 15% of the total score, that gap routinely outweighs any realistic difference in price.

Answer the question asked: the number one fail

Evaluation panels sink more bids for this single failure than for anything else. The evaluator has no discretion to reward material that does not address the question, however impressive it sounds.

Take a plausible facilities management question: "Describe your approach to mobilising this contract, including how you will maintain service continuity during the transition period."

Here is how most responses open:

> CleanBright Ltd is a leading provider of commercial cleaning and facilities services with over 20 years of experience across the public and private sectors. We pride ourselves on delivering high quality, cost effective solutions, and our experienced and dedicated team is fully committed to exceeding expectations from day one.

Forty nine words, zero marks. It is a paragraph about the bidder, and the question was about the mobilisation. No plan, no timeline, no risk, not one word on continuity. An evaluator scanning for scoreable content finds nothing to underline, and "leading provider" is not a fact anyone can verify.

Same question, same firm, rewritten:

> We will mobilise this contract over six weeks in three phases: due diligence and TUPE consultation in weeks one to three, parallel running alongside the outgoing provider in weeks four and five, and full service start with daily audits in week six. Our contract manager is on site from day one of mobilisation, not day one of service. We ran this programme on a contract covering 12 sites in 2024, transferring 38 staff under TUPE with zero missed service days in the first month.

Every sentence now contains something an evaluator can underline: a timeline, a phased method, a continuity mechanism in the parallel running weeks, named accountability, and a past example with numbers attached. A panel can defend a 5 for that answer in moderation without leaving the page.

The discipline behind the second version is mechanical, not creative. Paste the question at the top of your draft. Split it into limbs: "describe your approach" is one requirement, "including how you will maintain service continuity" is a second, and every "including" clause is an instruction rather than a suggestion. Mirror the buyer's vocabulary. Before submission, hand the answer and the scoring ladder to someone who did not write it and ask them to mark it cold.

Evidence beats adjectives

Every claim in a winning tender carries one of three things: a number, a named example, or a method. Anything else is assertion, and assertion tops out at a 3.

The upgrade is usually a straight swap.

• "An experienced team" becomes "a team that has completed 14 public sector mobilisations since 2022"
• "Robust quality assurance" becomes "fortnightly audits against a 40 point checklist, with scores reported to the client monthly"
• "Excellent response times" becomes "97% of reactive jobs closed inside SLA across our current portfolio, measured quarterly"

Adjectives fail because they are claims without collateral. No evaluator can write "awarded 5 because the bidder said they were excellent" in a moderation record. Numbers, names and methods hand the panel the raw material to argue your case when you are not in the room.

Run one test on every paragraph before you submit: could your nearest rival paste this sentence into their bid unchanged? If the answer is yes, the sentence differentiates nothing and is spending your word count on the panel's patience.

Bid response structure: allocate words to marks

Word and page limits are not formatting rules. They are the resource budget of the entire contest, and evaluators mark what is on the page, not what you ran out of room to say.

The maths is unforgiving. A 1,000 word limit with a 300 word introduction about your company history means 30% of the budget went on content that maps to no criterion and scores nothing. Repeat that across every question and a third of your bid is dead weight.

So allocate deliberately. A question worth 20% of the quality score deserves twice the drafting time and twice the evidence density of one worth 10%. Inside each answer, run the same structure every time: the direct answer in the first sentence, then method, then evidence, then risk and how you will manage it. Use subheadings that mirror the limbs of the question, because your evaluator is reading you as bid number eleven at 9pm, and marks are lost to fatigue as often as to weakness.

And use the space. A 3,000 character box answered in 800 characters tells the panel precisely how much thinking you did. Full limits packed with scoreable content is the target. Full limits packed with padding is worse than short.

Social value: a tenth of the score, decided in advance

On central government tenders, social value carries a minimum weighting of 10%, and plenty of councils weight it higher. It is also the section bidders throw away most often: vague community promises improvised in the final hour, scored a 2, contest lost by half a mark.

Social value answers fail exactly the way quality answers fail: adjectives instead of delivery plans, promises instead of measurement. Build a small set of deliverable, costed, measurable commitments before you ever bid, then tune them per contract. We cover what evaluators actually reward in the social value guide for SMEs, so this article will not re-explain it. The point that matters here: a tenth of the marks should never be a last minute improvisation.

The winning answer was written before the tender dropped

Here is the edge that never appears in lists of tender writing tips: the strongest bid on the table was mostly written before the tender was published. Not because that bidder writes fast. Because they started early.

A tender window runs four to six weeks. The buying decision took shape over the previous year: budget approval, market engagement, a specification drafted with somebody's strengths in mind. Suppliers who worked that period know the buyer's real problem, the incumbent's weak points and the likely evaluation emphasis before the ITT confirms any of it. Their case studies, mobilisation plans and method statements already exist, tuned to that buyer. For them the tender window is an assembly job. For everyone else it is a scramble.

That edge is research, not prose. Learn the buyer's pipeline, find out who holds the contract today, and work the six month pre-tender window while your competitors wait for an alert email. Buyer intentions are increasingly public: pipeline notices under the Procurement Act, published forward plans, and Crown Commercial Service commercial pipelines all signal what is coming.

Every one of those 1,567 awards carries an end date, and every end date is a future tender that somebody will start writing too late.

Frequently asked questions

What do tender evaluators actually score?

Your written answers against the published criteria and weightings, plus price through a disclosed formula. On quality questions they score what is on the page: relevance to the question, method, and evidence. Reputation, relationships and anything you meant to say are not scoreable.

What does a 5 out of 5 tender answer look like?

It answers every limb of the question directly, lays out method step by step, evidences each claim with numbers or named examples, and covers risk. Above all, it gives the evaluator quotable material to defend the top mark in moderation. Generic capability text caps out around a 3.

How long should a tender answer be?

As close to the limit as you can fill with scoreable content. Treat limits as a budget and allocate words in proportion to marks: cut introductions and boilerplate, and never leave a heavily weighted question thin while you polish a minor one.

What is the difference between MEAT and MAT?

MEAT, the most economically advantageous tender, was the award basis under the old regulations. The Procurement Act 2023 replaced it with MAT, the most advantageous tender, deliberately dropping "economically" to signal that quality and social value can be weighted hard. Expect quality led evaluations, not price races.

Can you win a tender with no public sector experience?

Yes, because evaluators score evidence, not badges, and strong private sector proof beats weak public sector proof question by question. You need transferable case studies, the compliance basics in place, and honest contract selection. Panels are required to score the answer in front of them, not the logo on top of it.

The next tender you write is already visible

The contracts you will bid on next quarter are taking shape now. There are 232 tenders open as you read this, and 1,567 recent awards ticking toward expiry with incumbents in place and buyers already planning what comes next. AtlasRevenue tracks all of it live across 28 sector desks: open tenders, expiring contracts, buyer behaviour. Run a scan and start writing your next winning tender six months before your competitors know it exists.