CAPTURE.
New public YouTube Shorts are observed as close to launch as the collection system allows. Publication time and first-observation latency stay visible. A late catch is never relabelled as T+0.
INDIA / SIGNAL DESK №01
TRUST PAGE / METHODOLOGY
MEASURE THE MOVE. PRESERVE THE EVIDENCE.
Clipped is not asking agencies to trust a black box. The method is designed to make timing, comparison, and uncertainty inspectable.
DECLARED PIPELINE
Capture → observe → normalize → score → freeze → verify.
NO HINDSIGHT
ALLOWED
ALLOWED
OBSERVE.
Snapshots are stored in declared cadence slots. Duplicate chains, sub-cadence observations, malformed jobs, and cross-platform leakage are treated as integrity failures—not extra evidence.
NORMALIZE.
Raw movement remains raw. Channel normalization asks whether the Short is ahead of its own expected curve. Peer normalization asks whether it is ahead of comparable creators at the same content age.
SCORE.
CVIE combines the declared evidence layers only when their required inputs are present. A missing comparison does not quietly become a zero, and incomplete rows do not become confident scores.
FREEZE.
A candidate call is timestamped before the outcome window closes. The explanation, client fit, confidence, and opportunity expiry are preserved as they were known at decision time.
VERIFY.
The later outcome is compared with the frozen call. Correct, incorrect, late, incomplete, and inconclusive cases remain distinct so the product cannot improve its history after the fact.
THE BOUNDARY
WHAT CLIPPED DOES NOT CLAIM.
Clipped does not guarantee virality, causally prove why a video moved, or claim that one platform predicts another. It produces an experimental decision signal whose value must be tested prospectively.