Assessment review cross-listing study, dataset 2026-09-02 (sharpassessment.com/assessment-review-cross-listing/).
One row per product and review site in data/signatures-<cohort>-2026-09-02.json: canonical URL, displayed review count and
one-decimal rating as read from the page's own structured data, listing name, sub-ratings where displayed as numbers, the
saved page's SHA-256 and byte size, UTC retrieval time, discovery step. data/discovery-trace-*.json records every step 1
and 2 search attempt (query, saved response files with SHA-256, the file used, the first accepted item, whether it
reproduces the recorded decision); data/profiles-probe-*.json every step 3 attempt; data/capture-*.json and
data/supplement-*.json the fetches; data/file-inventory-2026-09-02.json every saved page on disk with its status. The scripts are
the exact versions used. Saved pages are available on request from support@sharpassessment.com (their hashes are in the
rows). Rows whose saved body was a fetch-layer error response rather than a review-site page carry capture_status, http_status
and error_type and the status 'profile found in discovery; capture unresolved (<status>)'; they enter no eligibility count.
Capterra, GetApp and Software Advice rows carry listing_category and category_ok; a listing outside the HR category family
is labelled 'listing rejected' and counts as not found (amendment 6 in SCOPE.md).
Cohort 1 is the headline frame; cohorts 2 and 3 are sensitivity samples; the appendix enters no denominator.