Tender Enablement11 September 20268 min read

Architecture and Proposal Writing 101 — Part 4 of 12

Make it easy to award the point: requirements, assumptions and traceability

An evaluator with a scoring sheet cannot award a point they cannot find. Part 4 of the series covers answering requirements in the client’s own structure, separating facts from assumptions and commitments, and using a traceability matrix to catch gaps before an evaluator does.

#Tender Enablement#Proposal Strategy#Solution Architecture#Risk Management

Picture the person scoring your proposal. They have a spreadsheet with the RFP’s requirements down the left, a column for scores, several proposals to get through, and a deadline. For each requirement they need to find your answer, decide whether it meets the need, and write a score they can defend if another bidder challenges the result.

If your answer to requirement FR-012 is buried on page 47 inside a section called “Our Innovative Platform”, you are relying on a tired evaluator to go looking. Many will not. The point is not lost because the solution is weak. It is lost because nobody could find the evidence.

This part is about making the point easy to award, and about the discipline that sits underneath: being exact about what you are claiming, what you are assuming and what you are promising.

We are following one fictional pursuit throughout this series. Rivermark Water and Ridgeline Digital are fictional, and all figures are illustrative. Rivermark wants a customer self-service and case management platform; Ridgeline is bidding with a teaming partner and, in Part 3, learned that Rivermark’s billing system has no API for writing data.

Use the client’s structure, not yours

Rivermark’s RFP numbers its requirements. Functional requirements, describing what the system must do, start with FR. Non-functional requirements, describing how well it must do it, start with NFR. For example:

  • FR-012 Customers can report a fault and receive a reference number.
  • FR-031 Customers can see the status of their fault.
  • NFR-004 The service is available 99.5% of each month.

Ridgeline’s response uses exactly those numbers, in exactly that order, in the response format the RFP specifies. Where the story needs explaining at more length, the answer to FR-012 states the response in one line and points to the section with the detail. The evaluator never has to search.

Most RFPs ask bidders to state compliance for each requirement using fixed terms. The usual ones are comply (fully met as asked), partially comply (met with a stated limitation), do not comply, and sometimes alternative proposed (met in a different way, with an explanation). Be honest here. A “comply” that turns out to be partial damages trust in every other answer, and in a contract it becomes a promise.

Four kinds of statement

Every sentence in a proposal is one of four things, and mixing them up is where many disputes begin.

StatementWhat it meansRivermark example
FactSupported by an authoritative sourceThe billing system has no write API (written clarification answer)
AssumptionNeeded because material information is missingStorm-week fault reports stay under 30,000
RecommendationThe bidder’s proposed decisionRead balances through the read-only views rather than the nightly export
CommitmentA promise that becomes contractual if acceptedFault reporting and status live before 1 July, subject to the stated assumptions

The rule is simple. Never present a recommendation as a client fact, and never present an assumption as an unconditional commitment. “Rivermark’s billing views support real-time queries” is a fact claim. If Ridgeline only believes it, the sentence must say so.

The traceability matrix

A traceability matrix is a table that links each requirement to everything that answers it. Think of it as the thread that connects what the client asked for to what will be designed, built and tested.

Req IDRequirementResponseEvidenceSolution elementDelivery itemAcceptanceStatus
FR-012Report a fault and receive a reference numberComplySection 5.2Case management, portal, WhatsApp channelWP-03UAT-07Confirmed
FR-031See the status of a faultComplySection 5.3Case management, notifications, portalWP-03UAT-09Confirmed
NFR-00499.5% monthly availabilityPartially complySection 7.1Platform, monitoringWP-02NFR-T02Open: billing dependency

A few terms in the headings. A work package (WP) is a defined chunk of delivery work with its own outcome, covered properly in Part 10. UAT is user acceptance testing: the tests the client runs to confirm the work does what was agreed. Acceptance is the evidence that lets the client sign off.

A traceability thread, and the gaps it exposesONE REQUIREMENT, TRACED END TO ENDREQUIREMENTFR-012what was askedRESPONSEComplyour answerEVIDENCESection 5.2where to lookSOLUTIONCase,portal,WhatsAppwhat does itDELIVERYWP-03who builds itACCEPTANCEUAT-07how it is provenSTATUSConfirmedwhere it standsWHAT THE MATRIX EXPOSESREQUIREMENTFR-044 Monthlycomplaints reportno solution coverage: a point at riskSOLUTIONNothing designedor pricedREQUIREMENTNobody asked for itno requirement: cost with no scoreSOLUTIONChatbot on thearchitecture diagram
Traceability links what Rivermark asked for to what Ridgeline will design, build and prove. Read row by row, it also exposes the two classic gaps: a requirement nothing answers, and a component nobody asked for.

The matrix does two jobs. For the evaluator, it proves coverage. For the bid team, it exposes two classic gaps:

  • A requirement with no solution coverage. When Ridgeline built its matrix, FR-044, a monthly complaints report for Rivermark’s board, had no reporting component and no work package. It had simply been overlooked. Given the audit finding about complaints tracking, that was a point Rivermark cared about.
  • A component with no requirement. Ridgeline’s first architecture diagram included a chatbot. No requirement asked for one. It added cost and risk and could earn no points, so it moved to an optional section with its own price.

Notice NFR-004. Ridgeline marked it “partially comply” with a clear reason: the platform can be engineered for 99.5%, but customers checking a balance also depend on Rivermark’s old billing system, which Ridgeline does not control. Being open about it in the matrix is far better than committing to a number that someone else’s system can break. (The first draft did not say this, and Part 12 shows who caught it.)

Strong assumptions

Assumptions are not a sign of weakness. Every proposal needs them, because no bidder knows everything at submission. What matters is how they are written.

A useful assumption has six qualities:

  • Necessary. It is there because something material is unknown, not to shift risk for the sake of it.
  • Specific. It names the system, the data, the people or the quantity.
  • Observable. Anyone can tell whether it turned out to be true.
  • Owned. It says who must act.
  • Bounded. It promises no more than it has to.
  • Paired with its impact. It says what changes if it proves false.

Here is the weak version most proposals contain:

Rivermark will provide all required access.

It cannot be tested, nobody owns it, and when access is late the argument about what “required” meant has already begun. Here is the version Ridgeline wrote:

Rivermark will provide read-only access to the billing system’s customer and balance views for two named Ridgeline engineers by the end of week 4 of discovery. If access is more than five working days late, integration testing and the Bridge 1 go-live date move by the same amount. Production access follows Rivermark’s privileged-access process.

The anatomy of a strong assumptionWEAK“Rivermark will provide all required access.”STRONG, READ TOP TO BOTTOMTHE QUALITY IT ADDSRivermark will provide read-only accessOWNEDNames who must actto the billing customer and balance views,NECESSARYCannot price integration without itfor two named Ridgeline engineers,SPECIFICSays exactly who and whatby the end of week 4 of discovery.OBSERVABLEAnyone can check whether it happenedIf it is over five working days late,testing and the Bridge 1 date move to match.IMPACT IF FALSESays what changes if it slipsProduction access still followsRivermark’s privileged-access process.BOUNDEDPromises no more than it must
The weak version cannot be tested, owned or priced. The strong version, taken from Ridgeline’s Rivermark bid, carries six qualities, and each one gives an evaluator or a delivery team something concrete to check.

(Bridge 1 is the first stage of the transition plan, when fault reporting moves onto the new platform. Part 9 explains the bridge states.)

The strong version is also fairer to the client. It tells Rivermark exactly what to prepare and when, so a delay becomes a planning conversation rather than a surprise.

Never assume silently

Some assumptions are too important to leave buried in a list at the back. If any of the following are uncertain, they belong in the main response, the risk register and the price explanation:

  • Compliance, such as what POPIA obliges each party to do.
  • Data geography: where data is stored, processed, backed up and supported from. For Rivermark, where WhatsApp message data is processed.
  • Recovery: what happens, and how fast, after a failure.
  • API readiness: whether an interface really exists and does what is needed.
  • Data quality: whether the data is clean enough to migrate or rely on.
  • Volume: how much, how often and at what peak.
  • Licences: who buys them, how many and for how long.
  • Funding: whether the budget is approved.
  • Support ownership: who supports what after launch.
  • Third-party costs, such as SMS and WhatsApp message fees.
  • Approvals and acceptance: who signs off and on what evidence.

Each item on that list is a place where a silent assumption later becomes a dispute. An assumption that changes the price, the date or who carries the risk should be visible to the person approving the bid, not only to the person who wrote page 83.

Exclusions and client responsibilities

Exclusions clarify where the offer stops. They should not hollow it out. Ridgeline excluded the replacement of the billing system, which the RFP itself placed out of scope, and the redesign of Rivermark’s tariff structure. It did not exclude testing, security monitoring or data migration of open tickets, because a platform without them cannot be run safely, however low the headline price looks.

Client responsibilities are the things the client must do for delivery to succeed. For Rivermark they included providing subject matter experts for workshops, test data with personal information masked, timely access, approval of customer-facing content, and a named service owner. These belong in the delivery schedule and the risk model, with dates, not only in small print. A responsibility nobody has scheduled is a delay waiting to happen.

What to take from this part

Answer in the client’s structure and numbering. An evaluator cannot award a point they cannot find.

Know which kind of statement you are making. Facts, assumptions, recommendations and commitments must never be blurred.

Build a traceability matrix and read it both ways. It finds requirements with no coverage and components with no requirement.

Write assumptions that are specific, owned and paired with their impact. “All required access” is not an assumption; it is a future argument.

Put material assumptions and client responsibilities where decision-makers will see them. Not in the small print.

Next in the series: Features are not reasons: win themes, value and proof.

How CloudNala can help

We build and check traceability matrices with bid teams, reading them in both directions to find uncovered requirements and unrequested components, and we rewrite the assumptions list so every item is specific, owned and tied to its impact. It is detailed work, and it is often where the easiest points are won back.


Work with CloudNala

CloudNala helps organisations move from technology ambition to practical execution across cloud, AI, data, platform engineering and digital services.

Whether you are exploring AI, modernising your cloud environment, building a public-sector digital service, or turning an idea into a working MVP, we can help you shape the roadmap and deliver the next step.

Book a Bid Review or write to us at consult@cloudnala.co.za