Case Study Claim Evidence Extractor

Extract case-study claims, metrics, outcomes, client names, industries, timelines, before/after language, and evidence context from public case-study pages.

What it does

Extracts case-study claims, metrics, units, timeframes, client/industry context, before/after signals, and risky unsupported wording.

Every row is source-backed and schema-complete. Missing, blocked, or unmatched evidence produces diagnostics instead of invented values.

Uses

• Extract case study claims
• Build proof point inventory
• Find vague marketing claims
• Export metrics from case studies
• Review case studies for sales enablement

Input

Use authorized public HTTP(S) URLs or deterministic htmlInputs. Private-network targets are rejected and allowedDomains can narrow fetching.

• startUrls
• sitemapUrls
• htmlInputs
• allowedDomains
• maxPages
• maxDepth
• includeEvidence
• includeRawSnapshots
• requestDelayMs
• timeoutMs
• userAgentMode
• caseStudyUrls
• claimPatterns
• metricPatterns
• extractClientInfo
• includeScreenshots
• redFlagTerms
• maxCaseStudyPages
• maxClaims
• maxEvidenceRecords
• maxChargeUsd
• includeReport

Limits in maxCaseStudyPages, maxClaims, maxEvidenceRecords, timeoutMs and maxChargeUsd are enforced. Zero is a hard zero charge budget.

Dataset

• sourceUrl
• sourceType
• pageTitle
• recordType
• recordName
• issueCode
• severity
• evidenceSnippet
• recommendation
• canonicalUrl
• httpStatus
• scannedAt
• caseStudyUrl
• clientName
• industry
• claimText
• metricValue
• metricUnit
• timeframe
• beforeAfterSignal
• evidenceContext
• redFlag
• claimType

Reports:

• case-study-claim-evidence-extractor-summary-report.md
• case-study-claim-evidence-extractor-records.csv
• case-study-claim-evidence-extractor-issues.json
• case-study-claim-evidence-extractor-evidence-samples.json
• case-study-claim-evidence-extractor-run-metadata.json

PPE pricing

Platform-usage pass-through is off.

Event	Price
actor-start	$0.035
case-study-page-analyzed	$0.0115
claim-record-extracted	$0.0065
evidence-gap-detected	$0.0115
executive-report-generated	$0.4

The actor-start event is accepted before fetching. Source, record, signal, gap, and blocker events are deduplicated and charged before their paid output. Partial charges, event limits, and local caps stop later output. Reports are written only after the report event is accepted.

Safety and limitations

• Public evidence only; no login, CAPTCHA bypass, or access-control evasion.
• Output reflects source content at scan time.
• Attribution, claim permission, pricing interpretation, partner status, and launch decisions require human ownership.
• Store discounts, tasks, categories, icons, publication, and scheduling are separate gated live configuration.

Local verification

pnpm --filter @junipr/case-study-claim-evidence-extractor build
pnpm --filter @junipr/case-study-claim-evidence-extractor test
pnpm --filter @junipr/case-study-claim-evidence-extractor smoke