The fourth step of the loop is the one the category skips. Re-measurement uses the same questions and repeats as the baseline and a change rule fixed in advance — so a before-and-after is evidence, not a story.
Read the change rule See it on our own store
The analysis of our own intervention study was pre-registered on 11 September 2026 and cannot be edited after the data came in.
A platform that contradicts itself on 40% of repeats moves your one-sample score several points a day on its own. A two-point "improvement" measured once is indistinguishable from that.
If the question set is refreshed, expanded or re-picked by the tool, the second number is a different measurement. Provenance moved the same store by nineteen points in one week.
A threshold set once the numbers are in can always be set where the result clears it. A rule registered first cannot.
The baseline's question set, platforms, markets and repeat count are stored with the project. Re-measurement uses exactly those; new questions start a new baseline, never a comparison.
Two-sided Fisher exact test on cited-versus-not counts per platform, at the sample size the project has. The smallest detectable change is printed with the baseline so nobody expects a two-point move to register.
Thirty days for retainers, six weeks for the Sprint; monthly or fortnightly on the Standard plan. Same time of day where the platform allows it.
Per platform: changed, within noise, or new baseline. The report draws no arrow through a change the test did not support, and the CSV of both runs is attached.
Baseline and re-measurement drawn on the same track. If the bands overlap, the page says so before it says anything else.
Printed with every baseline: at 144 answers, about six points. The number you are told to expect before the work, not after.
Which lost questions were won, which were not, and the sources cited in each run — so the next five briefs are chosen from evidence.
Every run is labelled category, own-search-queries or natively authored. Two runs with different labels are never compared as before-and-after.
The intervention study's analysis script and its checksum were published before the data existed. The same code judges client re-measurements.
The print layout carries the interval, the test result and the plain-language verdict; an agency's name and colour where the brand goes.