Skip to content
Menu
← Operating guidesProduct & Technology · Operating guide

Abrasim operating guide

The Evidence Artefact Decision: Prototype, Proof of Concept, Concierge Test or MVP

Choose the smallest safe prototype, test or other evidence artefact that can answer your question, with clear acceptance and stopping rules.

By · Published · Updated

For
Product Leader, CIO, Founder
Leadership decision
Choose no-build, prototype, proof of concept, concierge, MVP, validation, change, defer or stop for one material uncertainty using a pre-agreed evidence threshold, permissions, safeguards and human decision gate.
Source cut-off
2026-08-26

Choose the simplest safe way to answer the question holding up your decision. That may be a prototype, proof of concept, concierge test or MVP, or it may require no build at all. Agree the permissions, evidence limits, acceptance criteria and stopping conditions before starting.

1. The Market Connection Context

The recognisable failure is a team asked to “build an MVP” before anyone has stated what remains uncertain. A clickable screen is used to claim demand, a technical proof of concept is shown to customers as if it proves usability, or a concierge test quietly depends on manual labour that will never exist in operation. Investment then follows artefact momentum rather than market, operating or technical evidence.

Accepted opportunity and investment boundary
→ one material uncertainty
→ decision and evidence threshold
→ permission and harm boundary
→ no-build, prototype, proof of concept, concierge test or MVP
→ observed evidence and disconfirming cases
→ proceed, validate, change, defer or stop

By [decision date], the Product Leader will assign NO-BUILD, PROTOTYPE, PROOF-OF-CONCEPT, CONCIERGE, MVP, VALIDATE, CHANGE, DEFER or STOP to one bounded evidence need, naming the decision it serves, participant and data permissions, evidence threshold, guardrails, owner and expiry. Selection authorises only the test boundary; it does not authorise product scale, portfolio priority or full delivery.

Use this guide after an authorised evidence handoff. That may be an accepted problem or demand decision under the Product Problem and Market Validation Standard, an approved route under the Build, Buy, Configure, Integrate or Stop Decision where investment diligence has reached that point, or a separately authorised evidence gate within a Product Opportunity Sprint. A build-or-buy decision is not an unconditional prerequisite when the sprint or problem-validation work has already identified an uncertainty that matters to the decision. Do not proceed if the problem or uncertainty, authority, ethics, resource limits or receiving owner is missing. This guide does not repeat broad opportunity validation, govern the portfolio under the Product Strategy, Roadmap and Portfolio Governance Standard, define full product operating standards or authorise delivery under the Product Delivery Governance Standard. Completing an artefact or receiving a favourable reaction does not prove an outcome.

Working files

The UK Government Service Manual describes prototypes as ways to explore, communicate and test ideas at differing fidelity rather than finished services (Making prototypes, VERIFIED official guidance). Its user-research guidance requires research to be planned around users, questions and ethical participation (Planning user research, VERIFIED official guidance). NIST's Secure Software Development Framework integrates security practices into software development rather than deferring them until release (Secure Software Development Framework, VERIFIED official framework). The UK Information Commissioner's Office explains how to incorporate data protection and privacy through the product design lifecycle, including purpose, necessity and safeguards (Privacy in the product design lifecycle, VERIFIED regulatory guidance). These sources support controls, not a universal artefact taxonomy. The method below is PROPOSED; its local permissions and threshold fit are REQUIRES VALIDATION, and any interpretation of observed behaviour remains INFERRED until challenged against retained evidence.

2. The Leadership vs. Delegation Boundary

The reader recognises that each uncertainty needs a deliberately bounded test, governance level, acceptance threshold and decision gate.

BoundaryRequired content
Product or investment leader ownsDecision question, risk appetite, resource boundary, allowed artefact class, evidence threshold, final proceed/change/stop disposition and operating consequence
Product, research or technical team may leadProtocol design, prototype or proof construction, recruitment, facilitation, observation, instrumentation, analysis, artefact production and knowledge transfer
SharedUncertainty statement, permissions, sample boundary, acceptance criteria, safeguards, measures, limitations and expiry
Cannot be outsourcedFinal business choice, research participation authority, privacy/security/legal approval, risk acceptance, source accuracy, customer accountability or operating ownership

An internal team can lead when it can challenge the requested artefact, has authorised access to participants and systems, and can stop the test without protecting sunk work. Independent support is justified when supplier incentives, founder attachment, power imbalance, disputed evidence or specialist research and security needs compromise challenge. A designer cannot infer market demand from preference feedback; an engineer cannot accept business viability; a product leader cannot waive privacy, safety or legal authority. Stop when consent is invalid, vulnerable people may be harmed, production access is uncontrolled, the answer is predetermined or the test cannot produce evidence that changes a decision.

For a handoff, retain the source decision ID, authorised boundary, receiving owner, evidence transferred, unresolved condition, resource authority, decision granted and not granted, next gate and expiry or reopen trigger. The Evidence Artefact Decision: Prototype, Proof of Concept, Concierge Test or MVP may receive a bounded The Product Opportunity Sprint: An Executive Standard for Build, Buy, Test or Stop Decisions or The Product Problem and Market Validation Standard: Proving Need Before Solution evidence gate before The Build, Buy, Configure, Integrate or Stop Decision: An Executive Technology Investment Playbook; investment, procurement, architecture, delivery and scale authority remain ungranted. Use The Product Problem and Market Validation Standard: Proving Need Before Solution as the anchor for problem truth, The Build, Buy, Configure, Integrate or Stop Decision: An Executive Technology Investment Playbook for delivery-route diligence and The Solution Architecture and Non-Functional Requirements Standard: Designing for Supportable Value for product solution architecture; reuse evidence only when its population, purpose and limits still fit.

Decision and escalation route

ConditionFirst routeEscalationRequired record
Missing consent, privacy, security or legal authorityPause the activity and refer to the qualified authorityStop or re-charter if the authority cannot be obtainedPermission and risk disposition
Disputed interpretation or evidence limitRetain both readings and commission independent challengeProduct Leader decides only within the stated evidence boundaryDissent, contradiction and disposition
Specialist method, technical or accessibility questionUse the named narrower specialist for that bounded questionReopen the charter if the answer changes the governing uncertaintySpecialist brief, method and limitation
Resource, operating or downstream scope changeReturn to the accountable owner for a bounded decisionEscalate to the executive owner before any new exposureResource, boundary and next-gate rows

Questions for evaluating a provider

  1. Which single uncertainty and decision threshold will your work address, and what will you explicitly leave unclaimed?
  2. How will you preserve source provenance, contrary cases, limitations, participant safety and reproducible handover evidence?
  3. What qualified authority must approve access, privacy, security, accessibility, legal and operating conditions, and how will a failed condition stop the work?
  4. How do you separate an artefact or prototype result from business outcome proof, and what would make you recommend validate, change, defer or stop?

3. The Step-by-Step Execution Standard

Step 01: Charter one evidence decision

Purpose and decision served Bound the decision before choosing or building an artefact.

Accountable owner Product Leader.

Delegated lead and contributors Product manager, researcher, technical lead, commercial owner and risk representatives.

Preconditions, inputs and access An accepted validation record under the Product Problem and Market Validation Standard, an investment boundary under the Build, Buy, Configure, Integrate or Stop Decision where that route has been reached, or an authorised evidence gate within a Product Opportunity Sprint. Known constraints and a receiving owner are required.

Actions

  1. State one decision, target population, context and decision date.
  2. Name allowed dispositions, excluded downstream decisions and resource ceiling.
  3. Record conflicts, independent challenge and stop authority.

Evidence to retain Signed charter, prior decision references, assumptions and conflict log.

Decision or scoring rule Proceed only when observed evidence can change one named decision.

Output and template Uncertainty charter and artefact decision register.

Acceptance gate CHARTERED, CHANGE or STOP.

Leadership checkpoint Product Leader confirms building is not the predetermined answer.

Human control point Named decision owner accepts scope and stop authority.

Failure modes and safeguards A feature request is not accepted as an uncertainty statement.

Change, stop or escalation condition Stop if no evidence could alter the intended action.

Handoff to the next step The charter controls uncertainty classification.

Step 02: Classify the governing uncertainty

Purpose and decision served Identify what is unknown and prevent one artefact from answering a different question.

Accountable owner Product Leader.

Delegated lead and contributors Research, design, engineering, operations, finance and customer owners.

Preconditions, inputs and access Accepted charter and prior evidence register.

Actions

  1. Classify the uncertainty as desirability, comprehension, usability, feasibility, viability, operability, control or adoption.
  2. State the claim, alternative explanations and disconfirming observation.
  3. Separate what is already known from inferred, proposed and unvalidated content.

Evidence to retain Claim contract, source references, alternatives and contradiction log.

Decision or scoring rule Select one governing uncertainty; secondary questions remain explicit and subordinate.

Output and template Uncertainty rows in the decision register.

Acceptance gate BOUND, VALIDATE, SPLIT or REJECT.

Leadership checkpoint Product Leader accepts which claim the artefact cannot establish.

Human control point Domain owners confirm source meaning within their mandates.

Failure modes and safeguards Positive reaction is not collapsed into demand, usability and viability.

Change, stop or escalation condition Split the charter when materially different uncertainties require incompatible tests.

Handoff to the next step The governing claim receives an evidence threshold.

Step 03: Define the evidence threshold and stop gate

Purpose and decision served Specify what observation would support, weaken or end the proposed path.

Accountable owner Product Leader.

Delegated lead and contributors Research lead, analytics, engineering, finance, operations and independent challenger.

Preconditions, inputs and access Bound uncertainty, authoritative baseline and decision options.

Actions

  1. Define eligible population, task or technical condition, observation unit and exclusions.
  2. Set support, validate, change and stop evidence conditions before execution.
  3. Record guardrails, sample limitations, expiry and attribution limits.

Evidence to retain Measurement contract, thresholds, owner approvals and dissent.

Decision or scoring rule No threshold may be rewritten after observing results without a versioned re-charter.

Output and template Permission risk and evidence threshold register.

Acceptance gate DECISION-FIT, CHANGE or STOP.

Leadership checkpoint Decision owner accepts disconfirming and stop evidence.

Human control point Measure and domain owners approve definitions and limitations.

Failure modes and safeguards Vanity activity and completion are excluded from outcome thresholds.

Change, stop or escalation condition Pause when no feasible observation can discriminate among options.

Handoff to the next step Thresholds define permission and risk needs.

Step 04: Establish permissions and harm controls

Purpose and decision served Ensure evidence collection is lawful, ethical, secure and proportionate.

Accountable owner Risk Executive.

Delegated lead and contributors Research, privacy, security, legal, data, accessibility and safeguarding owners.

Preconditions, inputs and access Protocol intent, participant population, data fields, systems and intervention risks.

Actions

  1. Identify consent, recruitment, privacy, accessibility, security and production-access requirements.
  2. Minimise data, privilege, exposure, duration and affected population.
  3. Define withdrawal, incident, deletion, compensation and mandatory stop routes.

Evidence to retain Approvals, participant materials, data map, threat assessment and incident plan.

Decision or scoring rule Missing mandatory permission cannot be offset by learning value.

Output and template Permission and risk rows.

Acceptance gate AUTHORISED, CONDITIONAL, REDESIGN or STOP.

Leadership checkpoint Product Leader accepts resource and scope changes required by safeguards.

Human control point Qualified owners approve within their mandates; participants retain withdrawal rights.

Failure modes and safeguards Prototype labels do not exempt a test from privacy, security or accessibility duties.

Change, stop or escalation condition Stop on invalid consent, uncontrolled production risk or material unmitigated harm.

Handoff to the next step Authorised boundaries constrain artefact selection.

Step 05: Select the minimum sufficient artefact

Purpose and decision served Choose the least burdensome test capable of reaching the accepted threshold.

Accountable owner Product Leader.

Delegated lead and contributors Research, design, engineering, operations, finance and risk owners.

Preconditions, inputs and access Bound uncertainty, evidence threshold, permissions, capability and resource ceiling.

Actions

  1. Choose NO-BUILD for evidence obtainable through existing records, interviews, observation or analysis.
  2. Choose PROTOTYPE for comprehension, interaction or workflow; PROOF-OF-CONCEPT for bounded technical feasibility; CONCIERGE for service behaviour and operating mechanism; MVP only for live integrated value and operability evidence requiring a real product slice.
  3. Record why lighter alternatives cannot reach the threshold and what the selected artefact cannot prove.

Evidence to retain Option comparison, rationale, rejected alternatives, burden and proof limits.

Decision or scoring rule Choose the lowest exposure and cost that can produce evidence relevant to the decision safely.

Output and template Artefact decision rows.

Acceptance gate NO-BUILD, PROTOTYPE, PROOF-OF-CONCEPT, CONCIERGE, MVP or STOP.

Leadership checkpoint Investment owner confirms the artefact does not imply scale funding.

Human control point Product, technical, operating and risk owners accept the selected boundary.

Failure modes and safeguards Build ambition cannot upgrade the evidence class automatically.

Change, stop or escalation condition Select no-build or redesign when existing evidence or a lighter test is sufficient.

Handoff to the next step The selected class informs a controlled protocol.

Step 06: Design the test protocol and sampling boundary

Purpose and decision served Make execution reproducible and interpretation appropriately bounded.

Accountable owner Research Lead.

Delegated lead and contributors Product, design, engineering, operations, analytics and participant support.

Preconditions, inputs and access Selected artefact, approved permissions and evidence contract.

Actions

  1. Define population, recruitment, scenario, task, environment and facilitation.
  2. Set observation method, event definitions, exclusions and disconfirming cases.
  3. Pilot the protocol without treating pilot observations as accepted evidence unless authorised.

Evidence to retain Protocol, recruitment record, script, sampling rationale and pilot changes.

Decision or scoring rule The sample supports only claims within its defined population and context.

Output and template Test protocol rows.

Acceptance gate READY, CHANGE, VALIDATE or STOP.

Leadership checkpoint Product Leader accepts the inference boundary.

Human control point Research and participant-authority owners approve protocol use.

Failure modes and safeguards Convenience participants are not silently represented as the market.

Change, stop or escalation condition Re-recruit or narrow claims when the eligible population is not represented.

Handoff to the next step The protocol defines artefact instrumentation.

Step 07: Build and instrument only the authorised boundary

Purpose and decision served Create enough fidelity to expose the uncertainty without accidental product commitment.

Accountable owner Technical or Design Lead, according to artefact class.

Delegated lead and contributors Engineering, design, operations, security, data and accessibility contributors.

Preconditions, inputs and access Approved protocol, scope, permissions, controls and evidence events.

Actions

  1. Build or configure only necessary interactions, mechanisms and evidence capture.
  2. Mark simulated, manual, live and omitted elements clearly.
  3. Verify access, logging, deletion, rollback, accessibility and incident controls.

Evidence to retain Artefact version, component boundary, test records, control checks and known defects.

Decision or scoring rule Every component must serve evidence validity, participant safety or a mandatory control.

Output and template Artefact and instrumentation rows.

Acceptance gate TEST-READY, CONDITIONAL, CHANGE or STOP.

Leadership checkpoint Product Leader rejects feature additions that do not change the evidence decision.

Human control point Qualified technical and control owners approve exposure.

Failure modes and safeguards Fidelity is not confused with validity or production readiness.

Change, stop or escalation condition Stop on uncontrolled access, misleading simulation or unreliable evidence capture.

Handoff to the next step The authorised artefact enters controlled execution.

Step 08: Execute and retain observations

Purpose and decision served Collect evidence without coaching, suppression or unsafe expansion.

Accountable owner Research Lead.

Delegated lead and contributors Facilitator, observer, technical operator, participant support and control monitor.

Preconditions, inputs and access Test-ready artefact, eligible participants, consent and live stop controls.

Actions

  1. Confirm eligibility, consent, context and participant understanding.
  2. Execute the approved protocol and record observation separately from interpretation.
  3. Log deviations, incidents, withdrawals, missing data and disconfirming cases.

Evidence to retain Participation record, observations, event logs, deviations and incident decisions.

Decision or scoring rule Only protocol-compliant observations enter the primary evidence set; exclusions remain counted and explained.

Output and template Test protocol observation and exception ledger.

Acceptance gate ACCEPT, CONDITIONAL, EXCLUDE or STOP.

Leadership checkpoint Product Leader receives material deviation and stop decisions.

Human control point Participants can withdraw; authorised monitors can stop unsafe execution.

Failure modes and safeguards Facilitator help and manual intervention are logged rather than hidden.

Change, stop or escalation condition Stop on harm, consent withdrawal, security event or protocol invalidation.

Handoff to the next step Controlled observations support bounded analysis.

Step 09: Analyse evidence and alternative explanations

Purpose and decision served Determine what the test establishes, weakens or leaves unresolved.

Accountable owner Product Leader.

Delegated lead and contributors Research, analytics, engineering, operations and independent challenger.

Preconditions, inputs and access Observation set, exclusions, protocol deviations, thresholds and prior claims.

Actions

  1. Compare eligible evidence with pre-agreed thresholds and guardrails.
  2. Segment by relevant context and examine failures, contradictions and missingness.
  3. Test alternative explanations and record inference and attribution limits.

Evidence to retain Analysis record, calculations, excerpts within permission, contradictions and dissent.

Decision or scoring rule Artefact completion, participant praise and technical execution do not establish an outcome beyond the tested claim.

Output and template Evidence disposition and learning transfer register.

Acceptance gate SUPPORTED, INFERRED, REQUIRES VALIDATION or REJECTED.

Leadership checkpoint Decision owner reviews adverse and ambiguous evidence before recommendations.

Human control point Evidence and domain owners confirm definitions and limitations.

Failure modes and safeguards Selective anecdotes cannot override contradictory retained observations.

Change, stop or escalation condition Commission a different test only when remaining uncertainty is material to the decision.

Handoff to the next step Bounded evidence reaches the decision gate.

Step 10: Make the evidence disposition

Purpose and decision served Convert the evidence into one named, limited next action.

Accountable owner Product Leader.

Delegated lead and contributors Investment owner, research, technical, operating, finance and risk owners.

Preconditions, inputs and access Evidence analysis, limitations, guardrails, dissent and resource boundary.

Actions

  1. Choose PROCEED, VALIDATE, CHANGE, DEFER or STOP.
  2. State which claim is supported, which is not, and what remains unknown.
  3. Record separate resource authority, conditions, owner, expiry and reopening evidence.

Evidence to retain Signed decision, evidence references, conditions, dissent and funding boundary.

Decision or scoring rule Only the named human decides; no metric, prototype or build status decides automatically.

Output and template Evidence disposition rows.

Acceptance gate PROCEED, VALIDATE, CHANGE, DEFER or STOP.

Leadership checkpoint Investment owner confirms the disposition does not imply portfolio priority or scale.

Human control point Qualified owners retain risk, privacy, security and operating mandates.

Failure modes and safeguards Conditional proceed is invalid without owned conditions and consequences.

Change, stop or escalation condition Stop if evidence, authority or safe ownership is absent.

Handoff to the next step The bounded decision and learning transfer to accountable owners.

Step 11: Transfer learning and close the artefact safely

Purpose and decision served Preserve reusable evidence while ending temporary exposure and ambiguity.

Accountable owner Product Leader.

Delegated lead and contributors Research, engineering, operations, data, security and knowledge owners.

Preconditions, inputs and access Signed disposition, retention rules and downstream owner acceptance.

Actions

  1. Transfer claim, evidence, limitations, decision and open uncertainty to named owners.
  2. Delete, anonymise, archive or revoke data, access and temporary systems as authorised.
  3. Label retained artefacts as evidence tools, not production products or outcome proof.

Evidence to retain Handover acceptance, retention and deletion records, access revocation and knowledge note.

Decision or scoring rule Closure requires accepted learning ownership and completed control obligations.

Output and template Learning transfer and closure rows.

Acceptance gate ACCEPTED, CONDITIONAL or HOLD.

Leadership checkpoint Product Leader confirms no unauthorised operational dependency remains.

Human control point Receiving and control owners accept explicitly; silence is not acceptance.

Failure modes and safeguards Temporary concierge labour or test infrastructure cannot become hidden production operation.

Change, stop or escalation condition Hold closure until access, data and owner obligations reconcile.

Handoff to the next step Accepted learning enters evidence-expiry review.

Step 12: Review evidence expiry and reopen deliberately

Purpose and decision served Prevent a bounded test result from becoming permanent unsupported truth.

Accountable owner Product Leader.

Delegated lead and contributors Evidence owner, product, operations, technical, customer and risk owners.

Preconditions, inputs and access Decision record, retained evidence, changed context, incidents and downstream observations.

Actions

  1. Reassess population, context, technology, risk, operating mechanism and evidence age.
  2. Compare downstream observations with the original claim and guardrails.
  3. Renew within limit, validate, change, retire or stop the claim and artefact path.

Evidence to retain Review record, changed assumptions, new contradictions and human disposition.

Decision or scoring rule Evidence transfers only within its accepted population, context, claim and validity period.

Output and template Evidence disposition and learning transfer register.

Acceptance gate RENEW, VALIDATE, CHANGE, RETIRE or STOP.

Leadership checkpoint Product Leader confirms whether a new uncertainty requires a new charter.

Human control point Outcome and control owners confirm evidence within their mandates.

Failure modes and safeguards Old test results cannot be relabelled as current market proof.

Change, stop or escalation condition Re-charter on material change in decision, population, context, permissions or risk.

Handoff to the next step The evidence cycle closes or begins a newly authorised decision.

4. The Evidence & Measurement Checklist

Preflight and authority

  • The Product Problem and Market Validation Standard: Proving Need Before Solution problem evidence, an applicable The Build, Buy, Configure, Integrate or Stop Decision: An Executive Technology Investment Playbook investment boundary or the authorised The Product Opportunity Sprint: An Executive Standard for Build, Buy, Test or Stop Decisions evidence gate are referenced. Evidence: uncertainty charter. Owner: Product Leader.
  • One decision and governing uncertainty are named. Evidence: charter. Owner: Product Leader.
  • No-build is an allowed disposition. Evidence: decision register. Owner: Investment Owner.
  • Independent challenge and stop authority are explicit. Evidence: authority record. Owner: Product Leader.
  • Downstream scale, portfolio and delivery decisions are excluded. Evidence: boundary statement. Owner: Product Leader.

Evidence access and quality

  • Claim includes population, context, unit and source. Evidence: claim contract. Owner: Research Lead.
  • Alternative explanations and disconfirming evidence are recorded. Evidence: contradiction log. Owner: Independent Challenger.
  • Thresholds are fixed before primary execution. Evidence: versioned protocol. Owner: Product Leader.
  • Exclusions, missingness and deviations remain counted. Evidence: observation ledger. Owner: Research Lead.
  • Artefact completion is not represented as outcome evidence. Evidence: proof review. Owner: Evidence Owner.

Method execution

  • Artefact class matches the governing uncertainty. Evidence: option comparison. Owner: Product Leader.
  • Lighter tests were considered before build. Evidence: rejected alternatives. Owner: Investment Owner.
  • Sampling and context bound every inference. Evidence: protocol. Owner: Research Lead.
  • Simulated, manual, live and omitted elements are marked. Evidence: artefact boundary. Owner: Technical Lead.
  • Observation remains separate from interpretation. Evidence: evidence ledger. Owner: Research Lead.

Governance, privacy, security and risk

  • Participant authority and withdrawal routes are approved. Evidence: consent record. Owner: Research Authority.
  • Data and privilege are minimised to the evidence need. Evidence: data map. Owner: Privacy Owner.
  • Security, accessibility and incident controls are verified. Evidence: control checks. Owner: Risk Executive.
  • Production exposure has named rollback and stop authority. Evidence: test plan. Owner: Technical Lead.
  • Mandatory permission failure cannot be offset by learning value. Evidence: gate register. Owner: Risk Executive.

Deliverables and acceptance

  • Charter, threshold, permissions, protocol and decision reconcile. Evidence: cross-reference review. Owner: Product Leader.
  • Artefact limitations are visible in the decision pack. Evidence: limitation statement. Owner: Evidence Owner.
  • Rejected artefact options retain rationale. Evidence: option archive. Owner: Product Manager.
  • Temporary data, access and operations have closure plans. Evidence: closure register. Owner: Technical Lead.
  • The Product Strategy, Roadmap and Portfolio Governance Standard: Outcomes Before Feature Commitments portfolio and The Product Delivery Governance Standard: From Discovery through Production Acceptance delivery remain unauthorised. Evidence: decision limitation. Owner: Investment Owner.

Measurement

  • Measures include population, denominator and exclusions. Evidence: measurement contract. Owner: Analytics Lead.
  • Behaviour, feasibility, operability and outcome measures remain distinct. Evidence: measure register. Owner: Product Leader.
  • Guardrails remain separate from favourable evidence. Evidence: risk register. Owner: Risk Executive.
  • Thresholds and expiry are human-approved. Evidence: review record. Owner: Product Leader.
  • Attribution limits and alternative explanations remain visible. Evidence: analysis note. Owner: Evidence Owner.

Decision and handover

  • Named human assigns proceed, validate, change, defer or stop. Evidence: signed decision. Owner: Product Leader.
  • Resource authority is separate from evidence disposition. Evidence: investment record. Owner: Investment Owner.
  • Conditions have owners, evidence and consequences. Evidence: condition register. Owner: Product Leader.
  • Receiving and control owners accept transfer explicitly. Evidence: handover record. Owner: Product Leader.
  • Dissent reaches the decision owner unchanged. Evidence: decision pack. Owner: Independent Challenger.

Review, change and stop controls

  • Consent withdrawal, incident and harm can stop execution. Evidence: stop procedure. Owner: Risk Executive.
  • Protocol changes are versioned and affect evidence eligibility. Evidence: change log. Owner: Research Lead.
  • Evidence expiry or context change can reopen the claim. Evidence: review rule. Owner: Evidence Owner.
  • Old artefacts cannot imply current product readiness. Evidence: archive label. Owner: Product Leader.
  • Materially new uncertainty triggers a new charter. Evidence: review disposition. Owner: Product Leader.

Topic-specific measurement contracts

Measure A: Evidence-threshold coverage

FieldRequirement
Outcome or decisionDecide whether the selected artefact can resolve the governing uncertainty
MeasureEvidence-threshold coverage
TypeLearning
DefinitionMaterial claims with pre-agreed population, observation, threshold, disconfirming condition and owner divided by all material claims; report counts
Baselinebaseline required
SourcePermission risk and evidence threshold register
OwnerProduct Leader and Research Lead
CadenceBefore execution and at each accepted review gate
ThresholdNamed decision owner sets change or stop threshold before evidence collection
GuardrailComplete thresholds do not make the test valid or safe
AttributionProtocol, sample, context and concurrent research remain visible

Measure B: Eligible observation coverage

FieldRequirement
Outcome or decisionDecide whether the retained observations support bounded analysis
MeasureEligible observation coverage
TypeEvidence quality
DefinitionProtocol-compliant observations meeting eligibility and consent conditions divided by all attempted observations; report accepted, conditional, excluded and withdrawn counts
Baselinebaseline required
SourceTest protocol observation and exception ledger
OwnerResearch Lead
CadenceAt each evidence review
ThresholdResearch and decision owners set validate, narrow or stop condition before execution
GuardrailCoverage does not establish representativeness or causation
AttributionRecruitment, facilitation, missingness and context remain visible

Measure C: Governing uncertainty movement

FieldRequirement
Outcome or decisionDecide whether the test reduced decision-relevant uncertainty
MeasureGoverning uncertainty movement
TypeLearning outcome
DefinitionChange between the accepted pre-test and post-test proof state for the exact claim, with supporting and contradictory evidence retained
Baselinebaseline required
SourceEvidence disposition and learning transfer register
OwnerEvidence Owner
CadenceAt the evidence disposition gate
ThresholdProduct Leader assigns supported, inferred, validate or rejected state using the pre-agreed contract
GuardrailProof-state movement is not market or commercial outcome proof
AttributionPrior evidence, protocol quality and external changes remain visible

Measure D: Permission and safeguard exceptions

FieldRequirement
Outcome or decisionDecide whether execution remains within authorised participant, data and system boundaries
MeasurePermission and safeguard exceptions
TypeRisk guardrail
DefinitionCount and severity of consent, privacy, security, accessibility, production-access, withdrawal or incident deviations by population and period
Baselinebaseline required
SourcePermission register and exception ledger
OwnerRisk Executive
CadenceDuring execution and at every incident or review gate
ThresholdQualified owners define mandatory pause or stop conditions before execution
GuardrailAbsence of logged exceptions is not proof of no harm
AttributionReporting access and monitor capability remain visible

Measure E: Decision consequence coverage

FieldRequirement
Outcome or decisionDecide whether evidence produces an accountable next action rather than an orphaned artefact
MeasureDecision consequence coverage
TypeGovernance
DefinitionMaterial evidence dispositions with named owner, bounded action, resource authority, conditions, expiry and reopening rule divided by all material dispositions; report counts
Baselinebaseline required
SourceEvidence disposition and learning transfer register
OwnerProduct Leader
CadenceAt each decision and transfer gate
ThresholdInvestment owner sets hold or escalation condition after evidence review
GuardrailAdministrative closure does not prove decision quality
AttributionPortfolio, capacity and governance changes remain visible

Measure F: Accepted target outcome

FieldRequirement
Outcome or decisionDetermine whether later authorised product activity affects the accepted outcome
MeasureAccepted target outcome
TypeCommercial outcome
DefinitionEligible outcome units meeting the accepted definition divided by the eligible population in the accepted period
Baselinebaseline required
SourceAuthoritative customer, operating and finance records
OwnerOutcome Owner and CFO
CadenceAt the separately authorised outcome review cadence
ThresholdNamed product or investment authority sets change or stop threshold after baseline validation
GuardrailThe evidence artefact alone cannot be credited with later outcome movement
AttributionMarket, product, service, pricing and operating changes remain visible

Evidence classifications used in this guide

  • VERIFIED: authoritative evidence supports the fact within its stated boundary.
  • INFERRED: interpretation remains distinguishable from retained observations.
  • PROPOSED: an artefact choice or method awaits accountable acceptance.
  • REQUIRES VALIDATION: a material claim, permission or threshold is not decision-fit.

Controlled source record

SourceExact useApplicability limitCurrency and review trigger
GOV.UK Making prototypes; GOV.UK planning user research; NIST SSDF; ICO privacy in the product design lifecyclePrototype purpose, ethical research planning, secure development and privacy-by-design considerationsGeneral guidance; not proof that an artefact is safe, effective, compliant or decision-fit in this organisationRecheck source editions and local permissions before each evidence gate or when law, participant risk or processing purpose changes

Use the method with AI

A prompt to help you work through this guide

Use this prompt to organise your analysis. Share only evidence you are authorised to provide to that model. A person must remain responsible for the decision, privacy limits and final action.

Copy and adapt

You are helping me apply the Abrasim operating guide: “The Evidence Artefact Decision: Prototype, Proof of Concept, Concierge Test or MVP”.

Situation: Teams build the wrong artefact for the uncertainty they need to resolve and mistake output for evidence.
Decision to make: Select the lowest-exposure, least burdensome evidence intervention that can safely discriminate among next actions, including no build, without allowing artefact momentum to substitute for outcome evidence.
Scope boundary: One bounded evidence-artefact decision covering uncertainty classification, proof and stop thresholds, participant and system permissions, minimum sufficient artefact selection, protocol, sampling, instrumentation, execution, analysis, disposition, safe closure and evidence expiry.

Use only the evidence I provide. Do not invent facts, benchmarks, sources, outcomes, customer research or legal conclusions. Keep VERIFIED, OBSERVED, INFERRED, PROPOSED and UNKNOWN separate. Ask focused questions when the evidence or authority is missing.

Help me work through this method by returning:
1. An evidence table with claim, source, date, confidence and limitation.
2. The strongest competing explanations, including what would disprove each one.
3. Decision options with trade-offs, dependencies and the cost of waiting where evidence supports it.
4. The next smallest evidence-gathering action, its accountable owner and acceptance test.
5. A proceed, change, validate, defer or stop recommendation that a named human decision-maker can review.

Treat the attached Abrasim working pack as a blank working record. Do not fill missing rows with plausible data. Flag privacy/POPIA, security, accessibility, safety and authority issues. Cite primary sources for external claims and show uncertainty. End with the conditions that would make us change course or stop.

My organisation-specific evidence and constraints:
[Paste evidence, links, dates, decisions and limitations here]

Working rule: AI can help organise evidence and expose questions; it cannot establish truth, approve risk, make the business decision or replace qualified legal, privacy, security or accessibility review.

Next decision

Choose no-build, prototype, proof of concept, concierge, MVP, validation, change, defer or stop for one material uncertainty using a pre-agreed evidence threshold, permissions, safeguards and human decision gate.

Use the product engineering decision route when uncertainty crosses market evidence, interaction, technical feasibility, service operation, controls and investment; an internal team can lead when it has participant authority, protected source access, independent challenge and named human decision ownership.