Perspective / Strategy & transformation
Procuring digital delivery: acceptance evidence, residual risk and exit planning

A digital delivery contract needs a defensible route from business intent to acceptance of working capability. Detailed features will change as discovery exposes constraints, but evidence, decision authority and operational obligations should not remain vague. Commercial buyers can improve procurement by agreeing how those matters will be governed before comparing supplier promises. This article sets out an illustrative acceptance model for a customer portal, drawing on current public-sector guidance as a reference while keeping the decisions applicable to commercial organisations.
The executive view
- Connect business outcomes to requirements, tests, evidence and named acceptance owners.
- Separate functional delivery, operational readiness and eventual commercial benefit.
- Specify change decisions, residual-risk acceptance and executable knowledge transfer.
Procure a delivery problem with visible boundaries
Describe the process that must work, the people it serves and the commitments it can create. Include the existing systems, data dependencies and operating constraints that a supplier will encounter. A polished requirements document is less useful when it conceals an unavailable interface or an unresolved decision about customer identity.
Separate known requirements, assumptions and questions requiring discovery. Ask suppliers to identify which commitments depend on each assumption. This makes bids easier to compare and reduces the chance that the lowest apparent price simply excludes work another supplier has included.
The Consultancy Playbook, refreshed in November 2026, provides a current reference for setting up engagements and transferring knowledge. Its government-specific requirements do not automatically apply to a private buyer. The useful commercial principle is to define the capability and responsibilities being purchased, rather than rely on a broad promise of transformation.
Give every bidder the same material context. Identify interfaces the buyer controls, dependencies requiring another supplier's cooperation and constraints that discovery must resolve. Ask each supplier to state exclusions and assumptions in a comparable schedule. Otherwise a fixed price may conceal different meanings of integration, migration or operational readiness.
Separate the procurement decision from the final solution design. Early market engagement can reveal that a requirement is expensive or framed around an existing product rather than the underlying need. Preserve a common outcome and evaluation basis while learning about viable approaches. The resulting brief should make trade-offs discussable: for example, whether the buyer values an earlier restricted release, broader initial scope or stronger transition support within the available budget.
Translate outcomes into evidence that can be accepted
An outcome such as “customers can manage their accounts” needs a chain of more precise statements. Identify the relevant journeys, authoritative data, permitted actions and failure behaviour. Specify how each will be demonstrated and who has authority to accept the result.
Keep functional acceptance distinct from operational acceptance. A user completing a successful journey does not establish recoverability, support readiness or correct behaviour during a downstream failure. Both sets of evidence matter, but they may be owned by different people.
Create a traceability matrix connecting the business outcome, requirement, test, evidence location and accepting owner. The matrix should be proportionate to the project, not an administrative exercise detached from delivery. Its purpose is to make omissions and unresolved dependencies visible before a release creates real commitments. Include the treatment of agreed exceptions and the person permitted to accept their consequences.
Write acceptance statements that can fail. “The portal is secure” is too broad to evaluate; “a user from customer A cannot retrieve customer B's order through the supported API and export paths” points towards observable evidence. Add the agreed test scope and who reviews exceptions. The technical team should explain what the test establishes and what remains outside it.
Define the evidence format before delivery. A demonstration recording, repeatable test result, reconciliation report and practical operator exercise serve different purposes. Specify where approved versions are retained and how an evidence item maps to the release being accepted. A test from an earlier configuration may no longer cover the delivered system. Traceability is useful when it preserves those relationships, not when it merely produces a large spreadsheet of unchecked requirements.
Worked example: acceptance of a customer self-service portal
Consider an illustrative portal connected to a CRM and an order platform. The buyer wants fewer avoidable service enquiries, but acceptance cannot depend solely on a future enquiry-volume reduction that the supplier does not fully control.
| Requirement | Acceptance evidence | Accountable buyer role |
|---|---|---|
| Display the correct account’s orders | Representative access and identity tests | Product owner |
| Submit a permitted address change | Validated request and downstream confirmation | Operations owner |
| Handle an unavailable order service | Tested pending/error state and recovery | Service owner |
| Preserve agreed data relationships | Reconciliation against approved records | Data owner |
| Support and transfer the service | Runbook exercise and access inventory | Receiving operator |
Commercial benefit remains part of the business case, with an agreed observation plan. Delivery acceptance concerns evidence that the contracted capability is working under defined conditions. Separating the two makes accountability more precise without abandoning the intended outcome.
Extend the portal example to failures. The CRM can be available while the order platform is down; a customer can submit the same request twice; a support user can have access revoked during an export. Agree the expected behaviour in each case and rehearse it with representative data. A success-only demonstration will not expose how the service handles ambiguity.
Define migration acceptance separately from interface acceptance. The buyer should know whether historical records are complete, correctly linked and usable for the agreed purposes. Count reconciliation alone cannot establish that the right orders belong to the right accounts. For each material exception, record whether it is repaired, excluded under an approved rule or accepted with a documented operational response. The supplier and buyer should share the same meaning of complete.
Define non-functional requirements through business scenarios
Terms such as fast, secure and scalable need operational meaning. Identify critical scenarios, expected workloads, user conditions and dependencies. Agree how performance will be measured, which failures count against the service objective, and what happens when an external dependency degrades.
Select requirements according to business criticality and cost. A demanding recovery objective may require additional architecture, rehearsal and operational capability. Buyers should see that trade-off rather than receive an impressive target with no explanation of how it will be met.
A useful evidence pack contains workload assumptions, test design, results and known limitations. It should distinguish a controlled test from observed production performance. Accessibility, permissions and data handling also require defined scope and review. Do not accept a generic assurance statement where the actual business journey, its users and its operating environment remain untested.
Make workload assumptions commercially meaningful. An illustrative portal may need to support ordinary daily use and a concentrated month-end ordering period. Define concurrent activity, transaction mix, data volumes and representative devices before assessing performance. A load test using only cached read requests cannot establish the behaviour of write-heavy ordering or document generation.
Recovery requirements also need an exercise. A stated recovery time should identify what has been restored, which data may be missing and when the business can resume the agreed work. Include dependencies such as identity providers and external interfaces. Buyers can then compare the cost of the requested resilience with the consequence of interruption. The appropriate requirement is the one the organisation can justify and operate, not the most impressive number in a proposal.
Make change control preserve commercial clarity
An evolving backlog does not remove the need for a commercial baseline. Record the agreed objectives, dependencies, budget constraints and acceptance approach. When a change is proposed, assess its effect on delivery effort, operating obligations, risk and evidence.
Use a decision log that states what changed and why. A feature can be deferred while the intended outcome remains achievable; another change may alter the investment case or require a new contractual agreement. Those decisions should be made by named people with sufficient authority.
Avoid treating completion of a delivery ceremony as proof that the service is acceptable. Reviews, demonstrations and working increments are useful ways to expose progress, but the buyer still needs evidence about the agreed capability. Payment and acceptance arrangements should reflect the chosen delivery model and the specific commercial agreement. Legal and procurement teams should shape the final terms alongside technical and operational owners.
Use change examples during negotiation to test the mechanism. What happens if discovery shows that the legacy API cannot supply a required field, the buyer requests a new approval tier or a regulatory interpretation changes? Work through the decision authority, evidence, cost and schedule implications before one of those situations becomes urgent.
Keep commercial status visible at the backlog level. A prioritised item may still require approval if it alters the agreed scope or operating obligation. Conversely, an implementation detail may change without changing the contracted outcome. The delivery and procurement teams should use a shared decision record so that technical conversations do not accidentally create conflicting expectations. The final legal wording should reflect the actual delivery model and the responsibilities each organisation is prepared to accept.
Accept residual risk deliberately
Every delivery has limitations. The important question is whether the remaining risk is understood, owned and compatible with the proposed release. Record each material exception with its affected process, consequence, mitigation, review date and accepting authority.
Distinguish defects from deferred scope, unproven assumptions and operational constraints. Combining them into one issue count can disguise a critical concern among many minor observations. An apparently small unresolved identity defect may matter more than a long list of cosmetic improvements.
The current GovS 005 Digital standard is a useful public-sector reference for lifecycle responsibility and assurance. For a commercial programme, the practical application is a release decision supported by evidence and explicit accountability. It should be possible to explain why a risk was accepted, not simply point to a signed acceptance form.
Risk acceptance should be specific and time-bounded where appropriate. An exception might permit a limited customer cohort while a non-critical integration is completed; it should state the restricted scope and the condition for expanding it. An unresolved cross-customer data exposure is qualitatively different from a cosmetic limitation, so a total defect count is a poor release criterion.
Record who can accept each consequence. A product owner may approve a layout trade-off but lack authority to accept a material information-security exposure or a new financial obligation. Escalation paths should be usable within delivery timescales. This avoids the common failure in which the project reaches a deadline with nobody able to make the remaining decision. An acceptance signature should reflect an informed decision, with the evidence and limitations readily available.
Design knowledge transfer and exit from the beginning
Agree which repositories, accounts, environments, documentation, licences and operating knowledge the buyer needs to use or transfer the service. IP ownership, licence rights and access control are related but different matters. Having source code does not establish the ability to deploy or operate it.
A decision-ready procurement pack therefore includes the problem statement, assumptions, traceability matrix, non-functional evidence plan, change mechanism, residual-risk process and transition requirements. This gives suppliers a clearer basis for proposing delivery and gives buyers a stronger basis for comparing it. The objective is an engagement that remains understandable through implementation, operation and eventual change of supplier.
Include a practical exit rehearsal early enough to change the solution. Ask the receiving team to deploy from the documented repository, restore a representative backup and investigate a simulated dependency failure. The exercise can expose missing account ownership, private package access or undocumented configuration before those gaps become a transition crisis.
For an illustrative handover, a complete file inventory can still leave the buyer unable to operate the service if deployment relies on a supplier-owned account. Identify transfer or replacement steps for hosting, domains, certificates, secrets and licensed components. Preserve secure access rather than copying credentials into a handover document. Define the agreed transition assistance and evidence of completion in the commercial arrangement, with legal review of ownership and licence terms appropriate to the engagement.
A commercially useful supplier comparison
Compare bids through the same delivery scenario. Ask each supplier how it would prove the critical journey, migrate the relevant data, handle a failed dependency and transfer operational responsibility. Require assumptions and unresolved dependencies alongside the proposed price and schedule. This makes differences in approach visible without pretending every supplier must implement the same architecture.
Evaluate the buyer's own obligations too. Delayed access, unavailable subject-matter experts and unresolved policy decisions can constrain delivery regardless of supplier quality. Assign owners and decision dates for those inputs. An engagement becomes more predictable when both parties can see the conditions on which progress depends.
The commercial objective is an agreement that remains understandable under change. Acceptance evidence, residual-risk ownership and tested transition arrangements give the buyer a basis for making decisions throughout the lifecycle, while giving the supplier a clear explanation of what successful delivery must demonstrate.
Questions this raises
Can an agile project have clear acceptance criteria?
Yes. The criteria can define outcomes, critical journeys, non-functional expectations and evidence while detailed scope develops. Changes still need an agreed mechanism for assessing cost, dependencies, risk and the effect on acceptance.
What is the difference between delivery acceptance and benefits realisation?
Delivery acceptance establishes that an agreed capability works under defined conditions. Benefits realisation examines whether its use produces the intended business effect, which may also depend on adoption, demand and operational changes.
What should a software handover demonstrate?
That the receiving organisation has the rights, access, documentation and practical knowledge needed for agreed operating responsibilities. Exercises such as deployment and fault investigation reveal gaps that a document inventory alone can miss.
Sources & further reading
- The Consultancy Playbook 6 November 2026 (updated)
- Government Functional Standard GovS 005: Digital 1 October 2026 (updated)
From thinking to doing