Anatomy of a rule

All rules follow the same structure, regardless of what they change.

1. The action

What the rule changes. Three rule types ship: Bidding, Import and Negation — plus the word-level negation rule, which is a variant of the last. The action determines which type you are configuring; see the Rule types section, which also covers the state, budget and placement rules that are not generally available yet.

2. The criteria

The thresholds that must be true. Criteria are reusable: the same set can be shared between rules rather than retyped, and they live in their own library.

Some trigger modes need more than one Criteria. A Performance Deterioration rule carries a Historical Criteria as well, because it has to establish two different facts about the same term.

Rules may also carry optional Campaign Criteria. These read the total performance of the Campaign containing the target. When target Criteria and Campaign Criteria are both present, both must pass.

3. The trigger mode

How the rule decides its action is warranted. The trigger belongs to the rule, not to the action, and which modes are available depends on the rule type:

Rule type	Available trigger modes
Bidding	Standard Criteria (default), Loss Budget Guard, Momentum Guard
Negation	Standard Criteria (default), Performance Deterioration

An unnegation rule always runs Standard Criteria — the App forces it back, whatever else is set.

Other rule types do not offer a choice: they evaluate their Criteria directly, which is what Standard Criteria does.

4. The Data Window

How far back the rule may look. Default 90 days, hard maximum 180, with the most recent days ignored by default. See Data Window.

Two trigger modes read their own windows instead: Performance Deterioration scores its Historical Window, and Momentum Guard evaluates its Criteria over its baseline. On those rules the Data Window no longer decides whether the rule triggers — though on a Bidding rule it still drives the bid arithmetic itself, which is a reason to keep it short rather than matching it to the baseline. See Momentum Guard.

5. The scope

For the rules that create or remove negatives — negation, unnegation and the word rule: Individual ASIN or Product Group. The Import rule reads search-term evidence too, but has no scope setting; it is bounded by its search terms source instead. See Decision scope.

Scope widens the evidence, never the change.

Do not confuse this search-term scope with a Bidding rule’s Rule Applicability. Applicability filters targets by Ad Product, Billing Basis and Target Kind before Criteria are evaluated; it does not aggregate evidence.

6. The assignment

The assignment specifies which ad groups the rule applies to. Without an assignment, the rule has no ad groups to evaluate and will not make any changes.

Limits and boundaries

Separately from criteria, a rule carries boundaries it may not cross: how far a bid may move, and the floor and ceiling it must stay inside. Criteria decide whether to act. Boundaries decide how far the action may go.

A rule also carries a name and a priority.

Settings and schedule overrides

Data Window, recent-data exclusion, run weekdays and excluded dates come from the Rule Type Settings by default. A rule can override them without changing the shared defaults. The effective values are shown in the editor, so check whether a rule says Inherit or Override when two similar rules behave differently.

Cooldown: the countdown after a change

A Bidding rule carries a cooldown. Once it changes a target’s bid, that target is locked for a number of marketplace days and no Bidding rule may move it again until the lock opens.

The cooldown is a ceiling, not a fixed wait. A target that gathers evidence quickly gets a decision sooner; one that does not waits out the full countdown.

Why it exists

A bid change needs traffic before it means anything. Without a cooldown a rule would re-read the same target the next day, see the same thin evidence, and move the bid again — chasing its own tail.

The three settings

Setting	What it does
Cooldown days	The longest the target stays locked. Default 5.
Minimum clicks	Once at least one day has passed, this many clicks releases the lock early.
Minimum spend	Once at least one day has passed, this much spend releases the lock early.

Leave the two minimums blank and the target simply waits out the full countdown.

A State Change action ignores both minimums. Enabling or pausing a target is not a bid decision, so there is no evidence to accumulate — it waits the cooldown out.

Choosing a number

Set it from how fast the target actually gathers clicks. A high-volume keyword can settle a bid question in two days; a long-tail target may need a week, and a short cooldown there just means the rule keeps acting on noise. Setting the minimums lets you have both: a long ceiling for slow targets, released early for the fast ones.

What is not part of the rule

Delivery. What a rule decides and what reaches the account are different steps — the run produces a preview, and applying is what sends it.