Proof loop: one prompt, before and after
The tracked prompt is the report's #1 gap: "best customer analytics platform for PLG SaaS". After the fix shipped, the same prompt is re-tested daily on the same engines - the before/after lives in dated observations, not a slide.
Baseline citations0 / 12before the fix
First re-test1 / 12Early lift signal
Day 148 / 12held across 3 re-tests
Cadencedailypaid proof loop
Observation log
Baseline · day 0
0 / 12 observations 4 engines x 3 samples
Prompt returned rivals only; fix brief generated the same day.
Fix shipped · day 2
-
Answer block + schema deployed to brightmetrics.example. Re-tests scheduled daily.
Re-test · day 5
1 / 12 observations perplexity
First sampled citation appeared on one engine.
Re-test · day 9
5 / 12 observations perplexity, openai
Citation now sampled on two engines; confidence band still wide.
Re-test · day 14
8 / 12 observations perplexity, openai, google_aio
Held across three consecutive re-tests. Reportable per the ledger bar.
How to read this. Engine answers are stochastic, so every figure is a sampled observation count, never a certainty. Use as directional proof until the sample reaches stronger confidence. The public Proof Ledger only reports outcomes that clear the stricter aggregate bar.
Take RankEcho with you: Full walkthrough · Intro to RankEcho · How it works · Walkthrough PDF