Metric methodology

Performance Index · Q1.0

Post views divided by median views of preceding posts from the same channel, format family, source and reviewed creator relationship in the trailing 90 days. The scored post and future posts are excluded. Under eight preceding posts, results are early evidence. Images, carousels and ad-library records are not scored on views.

Saved results include the computation time, baseline, sample size and a reproducible 80% bootstrap interval for the median baseline. Current legacy view counts may be captured at different ages; a saved day-7 capture or estimate is recorded separately. A day-7 label never means historical views were observed at day 7 when they were not.

Identity and attribution

External post identity prevents duplicate counting. Provider, fetch time and payload hash describe provenance where available. Missing legacy provider versions remain unknown. Verified owned channels, registered mentions and matched advertiser identities provide evidence; hashtag-only finds and name-matched ad candidates stay outside brand numbers until a recorded human confirmation.

Blind attribution accuracy is shown privately to staff after sample review, not as an invented public accuracy percentage. Paid partnerships and affiliate content never enter organic creator typical.

Version history and remaining work

24 Oct 2026 · Q1.0 saved results

Current results are now populated for active workspaces. Main workspace indices read saved results; pending calculations are labelled. Evidence-based attribution replaces legacy ownership labels for active workspaces. Paused workspaces are unchanged.

Rolling out

Universal report stamping, equal-age score baselines, model-based answer disambiguation, the two-rater video golden set and complete AI-output checks remain unfinished. Nightly recomputation requires the updated application to be published.