Conduct the Evidence Audit

Competitive Decision Toolkit · Evidence Audit

Audit the wording of the claim, not the size of the source folder

An Evidence Audit decides what each evidence record is allowed to support inside a frozen decision. For every material claim, record the source, source role, date, evidence state, scope, directness, independence, conflicts, limitations and permitted wording. If the evidence cannot support the claim as written, narrow the claim, repair the evidence, or hold it out of the decision.

Decision rule: no material word enters the decision brief unless another reviewer can trace it to an attributable record and see why that record is permitted to support that wording.

Memorable explanation: an Evidence Audit is like checking the labels, expiry dates and seals on ingredients before cooking the decision. A full cupboard does not mean every ingredient belongs in this recipe.

Fast Action

Audit one material claim in three minutes

Write the exact claim. Do not start with a URL.
Attach the source and observed passage/data. Preserve what was actually seen.
Name the source role. Discovery, first-party publication, buyer/sales evidence, independent third party, modelled estimate, specialist evidence or internal record are not interchangeable.
Classify the evidence state. OBSERVED, INFERRED, ASSUMED, CONTRADICTORY or UNRESOLVED.
Test the eight controls. Attribution, currency, directness, scope fit, independence, corroboration, conflict and materiality.
Rewrite the claim to the evidence. Record the wording the evidence is actually allowed to support.

If you cannot preserve the source and observed basis: the next task is evidence repair, not stronger prose.

One canonical Evidence Audit Record drives every product surface

The article, browser inspector, workbook, PDF and handoff use the same governed structure. Layout may change; field ownership, controlled states and decision meaning may not.

Identity

Evidence ID · Decision ID · Claim ID · candidate/subject · criterion/decision use

Claim properties

Exact claim · MATERIAL or SUPPORTING · exact permitted wording

Provenance and evidence

Source reference · source role · source/checked dates · observed basis · evidence state

Eight quality controls

SOURCE ROLE FIT · attribution · currency · directness · context fit · independence · corroboration · conflict

Governance boundaries

Source-role-fit rationale · known limitation · collection/use boundary

Governance

Reviewer · reopen trigger · audit outcome · governed next action

Diagram explaining How to Conduct the Evidence Audit, including the key evidence checks, decision criteria and sequence described in the article. — Visual summary of How to Conduct the Evidence Audit, connecting the article’s evidence checks and decision criteria to the appropriate next action.

Source role answers “What kind of thing can this source prove?”

Do not rank source roles as universally strong or weak. Judge whether the role is competent for the claim being made.

Source-role-fit rule: record whether the source role is competent for the exact permitted wording: YES, NO or UNCLEAR. NO → HOLD / EXCLUDE. UNCLEAR → REPAIR / NARROW. Record the reason against the visible role table.

Source role	Good for	Does not prove by itself	Typical audit use
FIRST-PARTY PUBLISHED	What a company publicly says about its offer, pricing, policies or customers.	Market leadership, buyer preference or independent performance.	Observed product/positioning/contract claim.
BUYER / SALES EVIDENCE	Observed buying objections, alternatives considered and recent conversations.	Population-wide market share or universal buyer behavior.	Buyer relevance and substitution signals.
INDEPENDENT THIRD PARTY	Externally produced observation, benchmark or analysis when methods/scope are suitable.	Truth merely because it is external.	Corroboration and external context.
MODELLED / ESTIMATED	Directional estimates where method and uncertainty are understood.	Observed customer behavior or exact company facts.	Bounded inference only unless independently verified.
DISCOVERY / INDEX	Finding entities, pages, topics or candidate sources.	Commercial substitution, authority or decision relevance.	Discovery, never automatic proof.
INTERNAL OBSERVATION	First-hand internal records such as sales notes, implementation experience or win/loss evidence.	External generalizability without sampling/context review.	Decision-specific observed evidence.
SPECIALIST / AUTHORITY	Regulatory, technical, legal, security, financial or methodological requirements within competence.	Commercial conclusions outside that specialist domain.	Constraint or specialist-claim support.
DECISION RECORD	What was previously decided, by whom, under which assumptions.	Current market facts.	Audit trail and change/reopen logic.

Evidence state keeps observation, reasoning and uncertainty visibly separate

OBSERVED

The recorded source directly contains the fact, event, statement or measurement being cited within the relevant scope.

INFERRED

The conclusion requires reasoning beyond the observed record. The inference must stay labelled and bounded.

ASSUMED

The claim is being used without sufficient supporting observation. Material assumptions cannot silently become facts.

CONTRADICTORY

Material evidence supports competing versions of the claim. Preserve the conflict rather than choosing the convenient source.

UNRESOLVED

The available evidence does not yet support a defensible conclusion for the decision.

State is not “confidence.” It describes the relationship between the claim and the evidence. A carefully labelled inference may be useful; an unlabelled assumption may be dangerous even when it sounds plausible.

Eight quality controls determine whether evidence is usable for this claim

Materiality is a claim property, not a quality score. Known limitation and collection/use boundary are governance boundaries. These eight controls are the exact taxonomy used by the article, browser, workbook, PDF and canonical schema.

Control	Controlled state / question	Fail-closed condition	Repair
Source role fit	YES / NO / UNCLEAR — Is this role competent for the exact permitted wording?	NO or unresolved competence.	NO → hold/exclude. UNCLEAR → narrow or obtain role-appropriate evidence.
Attribution	TRACEABLE / NOT TRACEABLE	A material record is not traceable.	Recover provenance or hold the claim out.
Currency	CURRENT ENOUGH / STALE / UNCLEAR	A material condition may have changed.	Refresh, version or state the historical boundary.
Directness	DIRECT / INDIRECT	Indirect evidence is written as direct observation.	Relabel as inference or find direct support.
Context fit	IN SCOPE / PARTIAL / OUT OF SCOPE	Evidence is silently generalized beyond the Decision Contract.	Bound the wording or collect in-scope evidence.
Independence	INDEPENDENT / DERIVATIVE / UNCLEAR	Derivative repetition is counted as corroboration.	Trace shared provenance and collapse duplicate support.
Corroboration	PRESENT / ABSENT / NOT REQUIRED	A consequential inference lacks separate support.	Add independent support or narrow the inference.
Conflict	NONE KNOWN / MATERIAL CONFLICT	Material contradiction is hidden or presented as settled.	Open a governed conflict record and state what would resolve it.

Known limitation

Required free text: sampling, method, missing context, model uncertainty or another boundary downstream users must preserve.

Collection/use boundary

ACCEPTABLE / UNCLEAR / PROHIBITED. PROHIBITED evidence cannot be authorized by review; UNCLEAR must be repaired before use.

Audit depth should follow materiality, not document count

A decision-support audit should be proportionate. A minor descriptive claim may need one attributable, current, in-scope record. A material comparative, causal, financial, security or market claim may need stronger directness, independent corroboration, specialist review or an explicit conflict record.

Supporting claim

Confirm provenance, scope and wording. Do not spend senior-review time proving trivia that cannot change the decision.

Material observed claim

Require traceability, current-enough evidence, context fit and a clear limitation/reopen rule.

Material inference

Keep the inference label visible and require genuinely independent corroboration before the inference carries consequential decision weight.

Specialist or high-consequence claim

Use evidence and review appropriate to the domain. The Evidence Audit does not manufacture legal, security, technical or financial competence.

Five links can still be one source: trace evidence dependency

Corroboration is about independent origin, not the number of URLs. A press release copied by news sites, directories and AI summaries remains one underlying source chain unless another record adds genuinely separate observation.

Identify the earliest or primary origin you can trace.
Mark derivative pages as derivative rather than independent.
Preserve any transformation: excerpt, estimate, model, summary or analyst interpretation.
Count corroboration only when another origin independently observes a relevant part of the claim.

Practical consequence: deleting derivative repetition can make the evidence register shorter and the decision stronger.

Rewrite the claim until the wording matches the evidence

A high-quality audit often produces a narrower but more defensible claim. That is not failure. It is the product.

Overclaim

“Competitor A is winning enterprise buyers.”

Observed records

Enterprise-query visibility, competitor-authored case studies, a modelled traffic estimate and three recent sales mentions.

Permitted wording

“Competitor A is visible in enterprise discovery and has appeared in several recent sales conversations; the current evidence does not establish enterprise market leadership.”

Language rule: remove every adjective, comparative, causal statement or quantifier that the audited records cannot support.

Interactive · Client-side only

Evidence Audit Inspector

Audit one record against the visible method. The tool does not calculate a reliability score and does not submit your evidence to SERPIFAI.

Audit ID Decision ID Claim ID Candidate / subject Criterion / decision use Materiality Exact claim Source reference / URL Source role Source role fit Source-role-fit rationale Evidence state Source date Checked date Observed passage / data Attribution Currency Directness Context fit Independence Independent corroboration Conflict status Collection / use boundary Known limitation Permitted wording Reviewer / owner Reopen / expiry trigger

Visible reviewer confirmations The permitted wording does not exceed the observed evidence or bounded inference. Scope has been checked against the frozen Decision Contract. Known contradictory evidence has been preserved rather than omitted. Source-role fit was judged against the visible role table, not assumed from source prestige.

Complete the evidence record

The tool returns DECISION-USABLE, BOUNDED INFERENCE, REPAIR / NARROW, or HOLD / EXCLUDE.

No JavaScript? Use the source-role, evidence-state and eight-control tables above. The browser inspector adds convenience; the methodology remains visible in the article.

Fictional example: one claim, four source roles, one narrower conclusion

Claim: “Competitor A is winning enterprise buyers.” The following records are fictional and illustrate the audit method; they are not market findings.

Record	Role/state	What it supports	What it does not support
Enterprise-query visibility	DISCOVERY / INDEX · OBSERVED	Visibility for the sampled queries/date.	Buyer wins or enterprise market share.
Competitor-authored customer stories	FIRST-PARTY PUBLISHED · OBSERVED	The company publicly claims named examples.	Independent prevalence or leadership.
Third-party traffic estimate	MODELLED / ESTIMATED · INFERRED	Directional domain-activity estimate within method limits.	Exact enterprise demand or conversions.
Three recent sales mentions	BUYER / SALES EVIDENCE · OBSERVED	The competitor appeared in those conversations.	Population-wide preference.

Permitted conclusion

“Competitor A is visible in enterprise discovery and has appeared in several recent sales conversations; the current evidence does not establish enterprise market leadership.”

Contradictory evidence is a governed object, not an inconvenience

When credible records materially disagree, do not choose the source you prefer and do not average incompatible claims into false certainty.

Create or link a Conflict ID.
Record the competing Claim IDs / Evidence IDs.
State whether the disagreement is about fact, scope, time, method, definition or interpretation.
State what evidence would resolve the conflict.
Record whether the decision can proceed with the conflict visible or must wait.

Several sources that repeat the same underlying press release or dataset are not independent corroboration. Trace shared provenance before counting support.

Set the expiry rule before the evidence becomes stale

“Reviewed once” is not a permanent evidence state. Reopen a material record when a condition changes enough to alter the claim-source relationship.

The source is corrected, removed, versioned or materially updated.
A pricing, product, policy, buyer, geography or implementation condition changes.
New direct evidence contradicts the current permitted wording.
Shared provenance is discovered and apparent corroboration collapses.
The claim becomes more or less material to the decision.
The Decision Definition Canvas itself is reopened.

Audit evidence with a team without turning “review” into consensus theatre

Decision owner sets materiality. Analysts should know which claims could actually change the action.
Researcher preserves observation. Save the exact source, date and observed passage/data before interpretation.
Reviewer challenges source role and wording. Ask what the record can support, not whether it feels credible.
Specialists own specialist boundaries. Security, legal, finance or technical claims should be reviewed within competence.
Conflicts stay visible. A disagreement becomes a conflict record with a resolution condition.
Freeze before handoff. Changes to material claims or evidence states belong in the change log.

Portable handoff: what the next analyst should receive

Identity: Evidence ID, Decision ID, Claim ID, subject and decision use.

Claim properties: exact claim, materiality and exact permitted wording.

Provenance/evidence: source reference, source role, dates, observed basis and evidence state.

Eight quality controls: source-role fit, attribution, currency, directness, context fit, independence, corroboration and conflict.

Governance boundaries: source-role-fit rationale, known limitation and collection/use boundary.

Governance: reviewer, reopen trigger, audit outcome and next action.

The Excel HANDOFF is a true frozen-value snapshot: frozen values are stored separately from live source formulas. Any later governed-field drift produces DRIFTED / NOT READY.

Download Excel evidence-audit workbook Download PDF field guide

Ten mistakes that make an evidence register look stronger than it is

URL collection.

Links are saved without the claim or observed basis.

Source multiplication.

Ten pages repeating one announcement are treated as ten independent sources.

Role confusion.

A search index, review site or modelled estimate is allowed to prove a claim outside its competence.

Freshness by accessibility.

A live URL is assumed current even when the underlying fact is stale.

Observation/inference collapse.

An analyst interpretation is rewritten as an observed fact.

Scope transfer.

Evidence from another buyer, geography, product tier or time window is silently generalized.

Conflict deletion.

Contradictory evidence disappears because it complicates the recommendation.

Assumption laundering.

An unsupported assumption becomes a “finding” after appearing in enough decks.

Equal audit effort.

Minor descriptive claims get more review than decision-changing claims.

No expiry.

An evidence record has no trigger for refresh or reopening.

Where the Evidence Audit fits in the Toolkit

Governing frameworkCompetitive Decision Toolkit Required inputComplete the Decision Definition Canvas Later qualificationUse the Competitor Relevance Scorecard Later interpretationInterpret the Opportunity Matrix Decision handoffWrite the Executive Decision Memo Escalation boundaryWhen self-guided analysis is no longer sufficient

Questions about conducting an Evidence Audit

What is an evidence audit in competitor analysis?

It is a claim-level review that traces each material statement to its provenance, source role, evidence state, quality controls, limitations and permitted decision use before the claim enters a decision product.

How many sources make a claim reliable?

There is no universal number. Several derivative or out-of-scope sources may add no independent support. Audit source role, provenance, context and independence rather than counting URLs.

What is the difference between observed and inferred evidence?

Observed evidence is directly present in the recorded source within scope. Inferred evidence requires reasoning beyond the observation and must remain visibly labelled and bounded.

Can modelled traffic or SEO estimates be used?

Yes, when the method and limitations are understood and the wording stays directional. A modelled estimate should not silently become an observed buyer, revenue or market-share fact.

What should I do when sources conflict?

Preserve the contradiction in a conflict record, state the conflict type and identify what evidence would resolve it. Do not select a preferred source without explaining why.

When should a claim be excluded?

Hold or exclude a material claim when its provenance cannot be recovered, the evidence is prohibited to obtain/use, a material contradiction is unresolved, or the claim rests only on an unsupported assumption.

Does the Evidence Audit replace competitor qualification?

No. The Evidence Audit governs what the evidence can support. The relevance/qualification method determines which candidates belong in the competitive decision set.

When should the audit be reopened?

Reopen a material record when its source changes, expires, is corrected, conflicts with new evidence, moves outside the Decision Contract scope, or stops being material to the decision.

Research base, competitor-content gap and methodological limits

The method combines SERPIFAI's competitive-decision workflow with broader analytical-quality and evidence-governance principles. External sources support individual controls; they do not certify the proprietary Evidence Audit or prove a commercial outcome.

Source	Role in this guide	Limit
UK PHIA Common Analytical Standards	Auditable, fully referenced analysis; contradictory evidence; coverage, credibility, reliability, age/currency, uncertainty and discoverable audit trail.	Government intelligence standards are not represented as governing private competitor analysis.
GovS 010: Analysis	Managing uncertainty proportionately and making its decision implications visible.	Does not define SERPIFAI source roles or evidence states.
Framework for assessment of third-party evidence	Provenance, methods, uncertainty, bias, transparency, omitted evidence and dissenting results.	Domain-specific guidance is used for transferable evidence-quality principles only.
Center for Evidence-Based Management	Critical appraisal and use of multiple evidence sources in organizational decision making.	Evidence informs; it does not mechanically dictate the decision.
SCIP ethical intelligence guidance	Legal, honest and transparent competitive-intelligence collection boundaries.	Not legal advice and not an endorsement of SERPIFAI.

Competitor-content gap: mainstream competitive-analysis templates are useful for organizing competitors, pricing, features, SWOT and market observations. This guide's distinct job is narrower and deeper: govern the provenance and permitted wording of each material claim before that information can influence a decision.