A tender compliance matrix is the control list that connects the buyer's documents to your response. Every requirement gets a row, and each row names its source, where the answer will sit, what evidence proves it, who owns it and who has checked it. It is not a scoring shortcut and it is not a checklist you tick at the end. It is how an in-house bid team makes sure nothing in the pack is missed. That includes the mandatory certificate on page 40 of an appendix, the portal file-size rule and the third sub-part of a scored question.
Teams that skip the matrix still answer the questions, and still lose bids on things unrelated to writing quality: an expired accreditation, a missing declaration, an answer that covered two of the three things asked. A good matrix makes those gaps visible while there is still time to fix them, raise a clarification or decide not to bid.
This guide sets out a working method for UK public-sector style tender packs. You will get a stable column model you can use as a compliance matrix template, a way to keep conditions of participation separate from award criteria, the human review gates that stop a row being marked done too early, a hypothetical worked example, and guidance on where AI can help with requirement mapping and where it must not decide. Treat it as a practical method, not a procurement standard. The competition documents for each tender take precedence, and the rules differ between regimes (for example in Scotland, or for procurements still running under transitional arrangements).
What a bid compliance matrix is for
A bid compliance matrix does three jobs at once.
- It proves coverage. Every actionable requirement in the current document set appears once, with a source reference precise enough that another reviewer can find it in under a minute.
- It assigns accountability. Each row has one named owner and, for critical rows, one named reviewer. "The team" is not an owner.
- It shows true status. A row is only marked checked when a person has looked at the actual response and the actual evidence against the source. Everything else stays visibly open.
The matrix is an internal working tool, usually not shared with the buyer. It needs to be current, complete and honest about gaps. Build it the day the pack arrives, because that is when you find out whether you can meet the pass/fail requirements at all, which feeds straight into the decision to bid. Our bid/no-bid worksheet is designed to sit alongside it.
Conditions of participation vs award criteria
The most common structural mistake in a tender requirements checklist is putting every "must" into one mandatory bucket. Different requirement types behave differently, have different consequences and need different evidence. Keep at least these types distinct.
| Type | What it tests | Typical examples | Consequence of failure |
|---|---|---|---|
| Condition of participation (where the competition uses the concept) | Whether the supplier is capable and suitable to perform the contract | Legal and financial standing, technical ability, relevant experience | Tender can be disregarded |
| Other mandatory compliance or eligibility | Specific pass/fail requirements stated in the pack | Named accreditations, insurance minima, required policies, signed declarations | Often pass/fail or exclusion risk, as the pack defines |
| Award criteria (scored quality and price) | The merits of the tender against the published methodology | Method statements, mobilisation plans, social value, pricing | Lower score, not automatic rejection |
| Submission and administrative instructions | How the tender must be submitted | File types, word limits, portal fields, signatures, deadlines | Can make a response non-compliant or unreadable |
| Pricing and commercial schedules | The commercial offer in the buyer's format | Rate cards, pricing schedules, contract mark-ups | Errors can invalidate or distort the price |
| Delivery or contract obligations (optional) | What you commit to if you win | Service levels, mobilisation dates, reporting | Mobilisation and contract risk |
Why the distinction matters under the Procurement Act 2023
For procurements covered by the Procurement Act 2023, the Cabinet Office guidance on assessing competitive tenders is explicit that award criteria are distinct from conditions of participation. Conditions of participation check whether a supplier is capable and suitable to perform the contract. Award criteria assess the merits of the tender. Once a supplier meets the conditions of participation, its tender is assessed against the award criteria without regard to how it fared on those conditions. The same guidance states that a tender must be disregarded where the supplier does not satisfy the conditions of participation.
The guidance also says contracting authorities must publish the award criteria and the assessment methodology, or a summary of them, in the tender notice, with further detail in the associated tender documents if not published in full. For your matrix, that means the weightings, sub-criteria and any pass thresholds should come from the published documents, not from what you assume the buyer cares about.
Do not merge "other mandatory" into participation
A named accreditation or an insurance minimum is often pass/fail, but that does not automatically make it a condition of participation in the legal sense. Track it on its own type. This keeps your evidence list accurate, stops you arguing about categories under deadline pressure, and makes it easier to see which rows carry disqualification risk and why. If the pack is unclear about how a requirement will be treated, log it as Clarification pending and raise it before the clarification deadline.
How to build a tender compliance matrix step by step
Build the matrix early, before anyone starts drafting. The steps below assume one tender and one lot. For multi-lot bids, either run one matrix per lot or add a Lot column and filter.
Step 1: freeze the document set you are working from
Start with a short header block at the top of the matrix. Record:
- Notice or opportunity reference, and lot if any
- Submission deadline, with time zone
- Clarification deadline
- Bid lead name
- The exact document set and versions reviewed: ITT, specification, pricing schedule, draft contract, portal guidance, published clarifications and addenda
- The date and time the set was last checked
This sounds administrative. It is the part that saves you when the buyer issues a revised specification nine days before the deadline. When documents change, reopen every affected row. Do not assume last week's pack is still current.
Step 2: extract one actionable requirement per row
Work through the current pack and every published clarification. Split compound asks. "Provide your quality policy and evidence of its implementation" is two rows if both are required, because they may have different owners and different evidence.
Keep the wording close to the source. You are building a control list, not rewriting the tender. Paraphrase is where requirements quietly change meaning: "should" becomes "could", a list of three becomes a list of two.
Tag each row with a type from the table above, plus Clarification pending where needed. Use ID prefixes that reflect the type so the matrix reads at a glance. For example, P for participation, M for other mandatory, S for scored, A for administrative, C for commercial and D for delivery.
Step 3: use a stable column model
This is the core of the compliance matrix template. It works in a spreadsheet, a shared table or a structured document. Keep the columns stable across bids so your team learns one layout.
| Column | Purpose |
|---|---|
| Req ID | Stable ID you never reuse within the bid (for example P-02, M-07, S-12, A-03) |
| Source | Document, version, page and clause |
| Requirement (short) | One requirement, close to the buyer's wording |
| Type | Participation condition, other mandatory, scored, admin, pricing, delivery, clarification pending |
| Published limit or assessment detail | Word limit, format, weighting, pass threshold. Only if published |
| Response location | Question number, schedule, portal field or file name |
| Evidence needed | Certificate, case study, policy, spreadsheet, named person, CV |
| Owner | One named person |
| Reviewer | Named checker for critical rows and high-weight scored rows |
| Internal due date | Always before the buyer deadline, with time for review |
| Status | Not started, Drafting, Ready for review, Checked, Gap, N/A |
| Checked by and date | Who confirmed it, and when |
| Notes | Clarification references, addenda, risks, next action |
Published practitioner templates use similar structures. TenderSNAP's matrix, for instance, combines a tender record, one record per requirement with source, limits, response location, evidence, owner, reviewer and status, and a review handover section. BidSmithery's one-page version keeps seven core columns and moves detail such as certificate expiry dates into a separate evidence log. Either approach works. What matters is that the columns do not change halfway through a bid.
If you want a starting point, our compliance matrix example shows a filled-in matrix with requirement types kept separate, and lets you export the example or a blank template as CSV.
Step 4: map every row to location, evidence and owner
A row is incomplete if any of these three fields are blank:
- Where the answer will sit. Question number, schedule or portal field.
- What proves it. The specific artefact, not "see library".
- Who owns it. A named person with time to do it.
Leave gaps visible. A blank cell is better than a false "Done". Set internal due dates backwards from the deadline, with participation and other mandatory rows due first, because a gap there may change whether you bid at all.
Human review of compliance: the rules that keep the matrix honest
Human review of compliance is where most matrices fail. Not because nobody reviews, but because "Checked" gets used to mean "someone said it was fine". The status column only works if two statuses mean exactly what they say:
- Ready for review. The owner believes the response text and evidence are prepared.
- Checked. A named reviewer has checked the actual response and evidence against the current source row.
Agree these rules at kick-off.
- Participation rows. Do not mark a condition of participation Checked until the capability or suitability evidence the buyer asked for exists in the required form.
- Other mandatory rows. Do not mark them Checked until the certificates, policies, insurance proofs or declarations exist as specified. Check validity dates against the submission date, not today.
- Sequence the work. Do not begin heavy scored drafting until fatal participation and other mandatory rows are cleared, clarified, or explicitly accepted as a bid risk by the bid lead.
- Scored rows. Mark Checked only when the reviewer has confirmed every part of the question is answered, within the published limit, with the evidence it relies on.
- Final-file check. Re-check the final files against the matrix before submission. Around 24 hours before the deadline is a useful default if portal windows allow, and some published templates recommend the same.
- AI-assisted rows. If AI helped extract requirements or draft a first-pass matrix, a person still confirms every critical row against the PDF or portal text.
One person, usually the bid manager, owns the matrix as a whole. Requirement owners move rows to Ready for review, reviewers move them to Checked or back to Gap, and subject matter experts supply evidence without changing status themselves. When everyone can edit status, nobody is accountable for it.
Requirement mapping for scored questions
Scored questions deserve more granular requirement mapping than a single row. A question that asks for "mobilisation milestones, responsibilities and dependencies" is three things the evaluator will look for. If you map it as one row, it is easy to submit a strong answer on milestones and a thin one on dependencies.
A practical approach:
- Give the question one parent row with the weighting and word limit.
- Add a child row for each distinct element the question or published guidance asks for (S-03a, S-03b, S-03c).
- Map each child row to the paragraph or heading where it is answered.
- Note the evidence each element relies on: a plan, a named roster, a case study.
- The reviewer checks the child rows first, then the parent.
This is also where the published assessment methodology earns its place in the matrix. If the buyer publishes what a top-scoring answer looks like, record that in the assessment detail column so writers and reviewers work to the same standard.
Worked example (hypothetical)
The rows below are from an invented Lot 1 facilities management tender. They illustrate the method only and do not describe any real buyer, supplier or result.
| Req ID | Source | Requirement | Type | Response location | Evidence | Owner | Status |
|---|---|---|---|---|---|---|---|
| P-01 | ITT v2 §2.4 | Minimum annual turnover as stated for Lot 1 | Participation condition | Portal Q2.4 | Latest audited accounts | Finance lead | Checked |
| M-01 | ITT v2 §3.2 | Cyber Essentials certificate valid at submission | Other mandatory | Upload slot 3 | Current certificate PDF | IT manager | Gap |
| M-02 | ITT v2 §3.5 | Employer's liability insurance at stated minimum | Other mandatory | Upload slot 4 | Insurance schedule | Commercial lead | Ready for review |
| A-04 | Portal guide p.6 | Quality responses as PDF, max 20MB per file | Admin | All quality uploads | Combined PDF | Bid coordinator | Not started |
| S-03a | Spec §4.1 / Q3 | Mobilisation milestones | Scored (Q3: 20%, 2,000 words) | Q3, section 1 | Mobilisation plan | Ops director | Drafting |
| S-03b | Spec §4.1 / Q3 | Responsibilities during mobilisation | Scored | Q3, section 2 | Named roster | Ops director | Drafting |
| S-03c | Spec §4.1 / Q3 | Dependencies on the buyer | Scored | Q3, section 3 | None yet | Ops director | Gap |
| C-01 | Pricing schedule v1 | Complete all rate lines in buyer format | Pricing | Pricing schedule | Completed schedule | Commercial lead | Drafting |
What this tells the bid lead on day three:
- M-01 is a potential stopper. If the certificate will not be valid at submission, that decision needs making now, not the night before.
- S-03c has no evidence. The answer on buyer dependencies risks being generic. Someone needs to work out what the dependencies actually are.
- A-04 has no owner action yet, which is fine on day three but should be in the final-file check.
Under the rules above, S-03 as a whole is only Checked once a reviewer confirms all three parts are answered within the 2,000-word limit and the plan and roster are attached.
Common mistakes with a tender requirements checklist
- One mandatory bucket. Participation, other mandatory and admin rules all lumped together, so nobody can see which gaps are fatal.
- Paraphrased requirements. Short-hand that drops a qualifier or a sub-part.
- Done means drafted. Status updated by the writer, not checked by a reviewer against the source.
- Stale document set. Addenda and clarifications published but rows not reopened.
- Evidence assumed. "We have that certificate" without anyone checking the expiry date.
- No final-file check. The matrix says compliant but the uploaded file is the old version.
Where AI helps with requirement mapping, and where it must not decide
AI agents can take much of the repetitive work out of a matrix. They should not own its conclusions.
Useful, with human confirmation:
- First-pass extraction of requirements from a long pack into the column model
- Clustering duplicate or overlapping requirements across documents
- Flagging rows with missing owners, empty evidence or no response location
- Comparing an addendum against the previous document version and listing rows that may need reopening
Not for unsupervised sign-off:
- Declaring the bid "compliant"
- Inventing or assuming evidence that has not been located
- Deciding bid/no-bid solely from a model score
- Treating a chatbot paraphrase as the buyer's requirement
- Moving any row to Checked
In practice, a Compliance agent produces a draft matrix and a gaps list, the bid lead and reviewers confirm rows against the source, and a separate drafting agent works only from approved sources and leaves gaps open rather than filling them. If you run this in a platform such as Grok Bot, validate in your own environment how source references, access permissions and change history behave before you rely on them. You can see how these roles fit together in our example agent workflow, and how review of the drafted answers works in how to review AI-assisted tender answers before they go out.
The matrix also benefits from a well-governed evidence source. If certificates, policies and case studies live in an approved bid knowledge library with owners and review dates, the evidence column becomes a lookup rather than a search.
Checklist before you call the matrix live
- Document set and versions recorded, with last-checked date and time
- Conditions of participation (if used) listed separately from other mandatory compliance and from scored criteria
- Admin and submission rules listed as their own rows
- Every row has an ID, source reference and named owner
- Empty evidence fields visible, not hidden or pre-filled
- Scored questions split into child rows where they ask for several things
- Clarification and addenda process agreed, including who reopens rows
- Final-file check scheduled before submission
- Checked status reserved for human review of real artefacts
Frequently asked questions
What is a tender compliance matrix?
A tender compliance matrix is a table that lists every requirement in a tender pack, one per row, with its source reference, response location, evidence, owner, reviewer and status. It is an internal control that shows the bid team what has been covered, what is missing and who is responsible. It is usually not submitted to the buyer.
What columns should a compliance matrix template include?
At minimum: requirement ID, source, requirement wording, type, response location, evidence needed, owner and status. Most teams also add published limits or weighting, reviewer, internal due date, checked by and date, and notes. Keep the column model the same across bids so the team learns one layout.
What is the difference between conditions of participation and award criteria?
Under the Procurement Act 2023, conditions of participation test whether a supplier is capable and suitable to perform the contract, while award criteria assess the merits of the tender. Once a supplier meets the conditions of participation, its tender is assessed against the award criteria without regard to how it did on the conditions. Always check how the specific competition documents use these terms.
When should you build a bid compliance matrix?
As soon as the tender pack arrives, before drafting starts. Building it early shows whether you can meet pass/fail requirements, which feeds the bid/no-bid decision, and gives writers a clear map of what each answer must cover.
Can AI build a compliance matrix for you?
AI can produce a useful first-pass extraction, flag missing owners or evidence and compare addenda against earlier versions. It should not decide that a bid is compliant or mark rows as checked. A person needs to confirm every critical row against the buyer's actual documents.
How TenderAgents helps
TenderAgents works with in-house bid teams to map the process you already run, set up AI agents inside it in a workspace you own and control, and train your team to use them. For compliance, that means a Compliance agent that drafts the matrix and gaps list from your tender pack, with your people confirming rows, owning evidence and keeping the final review and submission. If you have not yet mapped how a tender moves through your team, start with mapping your bid process before you add AI agents.
We are not bid management software, we do not write bids for you and we do not guarantee wins. Agents never log in to portals or submit tenders. If you want to explore implementing AI agents inside your tender team, the service page explains how the work runs, and the controlled pilot guide shows how to test it on a contained scope first.