The sales discussion ends with a friendly agreement: the customer wants faster installation, an intuitive reporting screen and the ability to use a product in a challenging environment. None of those expressions says what counts as acceptable. If a project starts from the meeting transcript alone, different teams will interpret the same conversation differently. The gap will emerge in design validation, service acceptance or, worse, a field failure.
The objective is not to turn every remark into contract language. It is to retain the customer’s intent, distinguish requirements from preferences and assumptions, then obtain confirmation of an operationally verifiable baseline. A reliable conversion process combines attentive listening with technical clarification, formal review and a feedback loop to the original decision maker.

01Separate stated needs from solutions proposed in conversation
Customers often describe a desired solution when they are actually communicating a business problem. “We need a real-time dashboard” may reflect concern about late escalation rather than a requirement for instantaneous data refresh. Ask what decision the user must make, what event triggers that decision and how late information can become before it causes loss. Preserving the underlying purpose helps the team avoid overengineering an attractive but irrelevant feature.
Record the exact customer wording as source evidence, then write a separate interpreted requirement. Identify whether it is mandatory, a preference, a working assumption or an unconfirmed question. Add who said it, the affected contract or product and any qualification attached to the statement. Avoid editing away uncertainty in the interests of a tidy spreadsheet; unresolved ambiguity must remain visibly unresolved.
Treat a recorded conversation as input evidence, not as an automatically approved requirement. When several people from the customer organization speak, record which person is an informed user and which has contractual authority. In a major project, a technical workshop can produce useful drafts while a separate commercial acceptance establishes what the provider must actually deliver. The two steps should be linked but not conflated.
02Probe measurable conditions and explicit exclusions
A testable requirement identifies the output, applicable conditions, acceptance method and responsible decision maker. For a service response promise, specify the incident classes, response event, calendar, operating hours, exclusions and evidence source. For a product requirement, specify configuration, operating environment, critical characteristics and test method. Terms such as robust, user-friendly or rapid are useful starting points, not objective acceptance criteria.
Clarification should also address what the customer does not require. A field installation in one region may need local-language documentation but not a translated user interface. Confirm the distinction before committing staff and software capacity. Record assumptions about customer-provided infrastructure, access, data and approvals, because delivery feasibility can depend on conditions outside the supplier’s control.
A useful test is whether a competent colleague unfamiliar with the conversation can determine pass or fail from the requirement record. If interpretation still depends on a particular salesperson being present, the specification is insufficient. Specify observable conditions, tolerances where justified, and agreed test circumstances; preserve uncertainty explicitly where measurement has not yet been authorized.
Standards context. ISO 9001:2026 defines quality-management-system requirements for reviewing customer requirements and operating processes. A requirement register and particular interview questions are implementation choices rather than universally prescribed ISO forms. Read the official ISO publication information.
03Build a traceable requirement record across channels
Requirements may arrive through workshops, emails, tenders, messaging platforms, service tickets and informal calls. Preserve a common identifier connecting the source, customer need, reviewed interpretation, owner, decision and verification evidence. One record should show which versions were superseded and why; copying the latest sentence into multiple departmental spreadsheets invites uncontrolled divergence.
A requirement register is an operating tool only if someone maintains it when the situation changes. Make the commercial owner responsible for obtaining customer clarification, while engineering or the service owner validates technical wording. Record the applicable document revision and any linked design, test or operational control. The user should be able to follow an acceptance result backward to the statement that authorized it.
Choose an identification system that will survive a product rename or change of software platform. A stable requirement identifier connected to version and source records is more important than an elaborate numbering convention. When customers have their own identifiers, retain them as an external reference instead of replacing them with internal codes that no one outside the provider recognizes.

04Verify internal feasibility before confirming the promise
Before sending a definitive acceptance, check the capacity, supplier dependencies, design rules, legal obligations and service resources that would be required to meet the interpreted requirement. The review needs enough cross-functional authority to resolve competing constraints; a salesperson cannot unilaterally waive a critical engineering or regulatory condition, and an engineer cannot silently change an agreed commercial scope.
Where verification reveals a gap, show options with their consequences: revised delivery time, a different product configuration, a conditional trial or a documented exclusion. Confirm the chosen option with the customer rather than relying on internal assumptions. If delivery is contingent on customer action, record the condition and provide a checkpoint before work proceeds into an irreversible stage.
An offer review should include commercial consequences without reducing feasibility to price. A new reporting obligation may be inexpensive to develop yet impossible to meet with the available historical data. A different technical configuration may satisfy the requirement but alter the maintenance burden. The customer needs to know the distinction before accepting an apparently equivalent solution.
05Agree acceptance evidence before building the output
Agree how the customer will know that the requirement is satisfied. Evidence may be factory acceptance testing, a witnessed demonstration, service transaction records, measured response time or a signed deliverable review. Acceptance criteria should cover important exceptions and interfaces, not merely ideal conditions. The test owner and customer acceptance authority must be identified before a disputed result occurs.
When a requirement is qualitative, agree a structured demonstration rather than inventing a precise number. For example, “easy to use” could be evidenced by representative users completing a defined task without assistance under agreed conditions. The acceptance record should capture the tested version, observations, known limitations and any deferred items, making disagreement about the result manageable.
Design the acceptance demonstration using representative conditions, not only a favorable scripted example. Include interface behavior, customer-provided inputs and exceptions that materially affect the outcome. When a complete acceptance test is impractical, agree the sampling basis and the treatment of residual uncertainty. The test should be reproducible enough for an independent reviewer to understand its result.
06Close the loop with change control and learning
A validated requirement is not permanently fixed. Site conditions, customer personnel, legal duties and product configurations change. Route new information through a controlled review that shows what is affected, who must approve it and which tests or resources need to change. The baseline should not drift simply because one project participant believes that a later conversation replaced the written agreement.
After delivery, compare warranty claims, service exceptions and customer feedback with the earlier interpretation. If complaints repeatedly concern an unstated assumption, revise the organization’s discovery questions and sales-to-delivery handoff. Effective learning is demonstrated by a different question being asked on the next opportunity, not merely by adding another mandatory field to the template.
Revisit discovery questions when requirements fail after delivery. A repeated class of unmet expectations is often evidence that the provider consistently asks about features but not intended use, peak load or the user group affected. Improve the upstream dialogue and the approval boundary; avoid responding only with more complicated complaint categorization after the customer has already suffered a poor outcome.
From conversation to acceptance
Begin a requirement conversion workshop with one actual customer interaction. Capture original phrases, distinguish outcomes from proposed solutions, resolve the highest-consequence unknowns and produce a single reviewed requirement baseline. Invite the technical and service owners to test whether they could execute the requirement without asking the salesperson for unwritten explanations. Assign a customer approval contact where the matter affects agreed scope.
At acceptance, verify a representative transaction or demonstration against the reviewed criterion. Record the version tested, relevant conditions, exceptions and decision. Feed any misunderstanding into the commercial discovery checklist; a controlled record is valuable only when it reduces future interpretation disputes.
A practical requirement record can use fields such as source statement, intended business result, controlled interpretation, operating condition, acceptance method, accountable reviewer, customer confirmation and current status. Do not require every field for every casual observation; use the full record where a commitment affects deliverable scope or risk. In a service scenario, connect the stated response requirement to classification, timestamp provenance, customer access assumptions and evidence from an actual ticket. In a manufactured-product scenario, connect the need to the drawing revision, relevant specification and approved inspection method. The common discipline is that another competent person should be able to reconstruct how the original customer statement became an authorized and testable requirement. That reconstruction also helps resolve disputes: when the buyer claims the supplier promised an outcome, both can inspect the preserved discussion, the clarification, the accepted baseline and the delivered evidence. A record cannot manufacture agreement where none occurred; it can expose the unresolved question before the organization incurs irreversible cost.
Practical implementation checklist
- Capture original customer language and identify unresolved questions
- Translate outcomes into conditions and verifiable acceptance criteria
- Obtain cross-functional feasibility review before making commitments
- Confirm the approved baseline and maintain change traceability
Closing perspective
Customer understanding becomes reliable only when the original intent, internal feasibility and customer-agreed acceptance evidence can be traced to one another. The register is not proof of good listening; a successfully demonstrated requirement and an informed customer decision are.
A good requirement review should let a customer confirm intent and let an operator demonstrate conformity without relying on private recollection of the sales conversation.