Evidence and periods

The AI Analyst and an Automation rule cut time differently, and reading one as though it were the other causes most of the surprises.

The Analyst compares a recent stretch against the stretches before it, so it answers what changed. An Automation rule evaluates its thresholds across one Data Window as a single block, so it answers what is true over this period — and a window holding both a good spell and a bad one shows you the average of the two.

Do not read one as the other. “This target is losing money over 90 days” is a rule’s kind of statement. “This target stopped converting two weeks ago” is the Analyst’s.

The same query, two scopes

The other axis is scope. A query you negated on one ASIN has no traffic there by definition — it was blocked. Read across the rest of its Product Group, it may be steadily producing orders.

At Individual ASIN scope: no evidence, no reason to revisit. At Product Group scope: an obvious candidate to unblock.

What this means in practice

Say which period you are quoting. “This target is losing money” is not a claim until it names the stretch of time it covers. Most disagreements about a target are disagreements about that.

A short period shows change, not level. It is good at telling you something moved. It is bad at telling you what normal was.

Widen the evidence before you widen the action. If a target does not have enough observations to settle a question, look at the group’s evidence. Do not lower your standard of proof instead.

How Vendlen applies this

Every Decision Brief states the period its evidence came from and the scope it applies to, as part of the brief. Every Automation rule carries a Data Window and, where it reads search terms, a scope. Both are recorded with the decision.