What is Rerank decay pattern?
The rerank decay pattern is the three-phase curve a previously cited chunk rides as rerank survival falls without intervention. Phase 1 — Freshness drift: chunk retrieves but the freshness signal stack has aged into the discount window; rerank survival compresses 25–40% over 8–12 weeks. Phase 2 — Claim erosion: chunk retrieves with current freshness, but competitor chunks have sharpened claim specificity on the same sub-query; rerank survival compresses another 20–35% as the cross-encoder picks the more specific claim. Phase 3 — Entity dilution: chunk retrieves on shape but the entity slot in the sub-query has shifted in the engine's entity-graph; rerank survival drops to near zero. Reading the decay phase per failing chunk is the operational discipline that scopes the fix — freshness drift gets a five-signal refresh, claim erosion gets a leading-sentence rewrite, entity dilution gets an entity-graph rebuild. Scoping the wrong fix for the phase produces no lift and burns editorial bandwidth.
How it relates to AI UGC
Visual rerank decay rides a parallel three-phase curve — Phase 1 (image freshness drift), Phase 2 (persona/composition erosion as competitor visual sets sharpen the carousel match), Phase 3 (visual entity dilution as the multimodal substrate re-weights the brand's visual entity). The visual decay window is materially shorter than the text decay window (4–12 weeks vs 8–12+ weeks for the text phases) which is why image refresh ships on its own queue. ppl.studio holds the visual side of the decay curve above the carousel threshold on persona-locked sets.
Key statistics
- Mid-2026 rerank decay phases compose: Phase 1 freshness drift compresses survival 25–40% over 8–12 weeks; Phase 2 claim erosion compresses another 20–35%; Phase 3 entity dilution drops survival to near zero (decay-phase audits, 2026).
- Phase 1 fixes recover 60–80% of lost survival; Phase 2 fixes recover 70–90% of the additional loss; Phase 3 fixes take 6–10 weeks to re-anchor — the recovery latency varies sharply by phase, which is why correct phase diagnosis is the load-bearing step in the fix scope (decay-recovery cohort, 2026).
- Programs reading decay forensics weekly catch Phase 1 inside the refresh window, Phase 2 inside the next sprint, and Phase 3 before it compounds — vs programs reading only aggregate citation share, which intervene 12–18 weeks late on average (decay-detection latency audits, 2026).