Resources · 61

Project acceptance: turn needs into criteria and evidence

Prepare scenarios, expected outcomes, evidence and exception decisions to assess a delivery through tasks actually tested.

· 4 min

Empty cyber intelligence operations room at night AI-generated editorial illustration

What this guide helps achieve

  • Start with a task
  • Write a testable expectation
  • Connect evidence to the scenario
  • Resolve exceptions

Quick check

  • Who should accept delivery?
  • Is an exception a pass?
  • Can tests be reused after a correction?

Step-by-step method

  1. 01

    Start with a task

    State who needs to accomplish what, in which context and under which constraints. The proposed method draws on GOV.UK service-assessment preparation without treating private acceptance as public certification.

    Deliverable: needs and priority journeys.

  2. 02

    Write a testable expectation

    Replace words such as fast or intuitive with an observable result and measurement context. Include errors, permissions, language and accessibility. A criterion states what the user obtains, beyond the presence of a feature.

    Deliverable: criteria and fictional test data.

  3. 03

    Connect evidence to the scenario

    Record version, environment, preconditions, action and outcome for each trial. A screenshot alone does not prove a journey succeeds. Prepare evidence a reviewer can inspect or replay without access to secrets.

    Deliverable: reproducible evidence record.

  4. 04

    Resolve exceptions

    Define blocking failures, permitted exceptions and owners before acceptance. An accepted exception needs a scope, reason and deadline; it does not turn a failed test into a passing test.

    Deliverable: exception register and decisions.

  5. 05

    Close with operations

    Test documentation, recovery and support handover too. For AI services, identify limits and ownership. Retain tested scope, open gaps and conditions requiring another acceptance review.

    Deliverable: delivery decision and exception follow-up.

Four situations for testing the method

Fictional teaching cases. Each situation connects a signal, a check and a decision bounded by evidence; none represents a client outcome.

Fictional situationConcrete checkDecision and limit
Close a project on a completed task — The plan marks a form as delivered, but its confirmation never reaches the authorised test recipient.Replay entry, validation, sending and receipt. Associate each failure with a step and an owner.Accept the interface and delivery chain separately; keep the business task open while its confirmation is missing.
Make a search requirement verifiable — A specification asks for relevant search without defining questions or expected results.Choose common, ambiguous and unanswerable queries. Specify useful results, filters and no-match behaviour.Accept against that shared set rather than a general impression of the screen. Treat a new intent as an explicit scope change.
A completed iteration must preserve the journey — A team quickly ships a new basket, but returning from payment loses the selected variant.Replay the complete journey before and after the iteration, including backtracking, errors and retrying. Keep the same expected outcome.Slice work around a usable task; a completed technical ticket does not close a journey regression.
Compare RFP responses on the same scenario — Two proposals answer the same request, but one includes migration while the other covers only building the site.Ask each bidder to describe migration of the same content, checks, training and operational handover. Record inclusions and assumptions.Align scope before comparing prices. Record exclusions that would change the choice without inventing a universal score.

Reusable worksheet

Complete with your authorised observations. These fields are a working template, not observed results.

FieldInformation to record
NeedPerson, task and context
CriterionExpected result and preconditions
TrialVersion, fictional data and observation
DecisionException, owner and deadline

Worked example

Illustrative situation

Fictional example: a form submits a case but loses its attachment on mobile.

Decision and expected evidence

Acceptance distinguishes a technical submission from a complete case; the need passes only when the recipient can retrieve the attachment.

Distinguish the mechanisms

MechanismPurposeCheck or limitation
DemonstrationUnderstand a selected journeyDo not treat it as full coverage
Business acceptanceVerify tasks and expected outcomesInclude errors and distinct user profiles
Technical reviewInspect contract and behaviourConnect outcomes to user needs

Management indicators

IndicatorWhat it measuresFirst action
Covered criteriaCriteria actually exercisedSeparate untested, passed and failed
Open exceptionsAccepted or blocking gapsName an owner and deadline
Replayable evidenceReconstructible scenariosRecord version and preconditions

Common pitfalls

  • Do not treat it as full coverage
  • Include errors and distinct user profiles
  • Connect outcomes to user needs

Frequently asked questions

Who should accept delivery?

The designated owner of the need, with input from required specialists. Name this responsibility before testing.

Is an exception a pass?

No. It remains a known gap; acceptance is a documented decision within a scope.

Can tests be reused after a correction?

Yes, with stable data and preconditions, naming the new version and potentially affected journeys.

Files: quarantine and withdrawal

Include late tasks in file acceptance tests

Acceptance tests should trigger withdrawal or replacement while analysis is running. Expected behavior concerns final state and actual access, beyond the task’s own success status.

  • Test failure, incomplete checks and retry.
  • Change version or policy during a task and verify rejection.
  • Withdraw access before a derivative returns, then replay permitted and forbidden reads.

View the three journeys and late task results →

Sources reviewed on · OWASP · File Upload Cheat Sheet · OWASP · Authorization Cheat Sheet · WHATWG · HTML input accept · PostgreSQL 18 · Transaction isolation

Official references

References consulted: . The method and worksheet propose checks to adapt to your context; they do not constitute certification.