Automation rules
An automation rule watches a metric and changes a setting when it crosses a line you drew. The nightly solve decides how to split the budget; automation decides when to move the dial you set it against — the group's budget, a campaign's target ROAS, a campaign's status.
Rules live on the group's Automation tab. Each one is a sentence: every day at a time, looking at a window, if a metric compares to a value, set a setting — for these campaigns, within these limits.
Two scopes
| Scope | Can change |
|---|---|
| Group automation | The group itself, or all eligible campaigns in it |
| Campaign automation | Only the campaign you opened it from |
Starting one
Start from a template for a ready rule you can change any part of:
| Template | What it does |
|---|---|
| Hold ROAS near target | When returns drift under the target you set, tighten the target so the platform bids more carefully and pulls ROAS back up |
| Buy traffic when ahead | When returns run well above the target, loosen it a little to win more impressions while they are cheap |
| Scale what is working | When a campaign beats its target and is spending real money, give it more daily budget |
Or start from scratch with an empty IF … THEN.
A rule has two editors and the same rule underneath: Sentence, which reads as plain English, and Detailed, which exposes everything. A few shapes have no sentence form — a condition that checks whether a metric exists rather than comparing it, a threshold that is a calculation, a metric that is itself a calculation, or an action set from a formula. Those stay in the detailed editor and the panel says why.
Time
How often it runs, when, and what it looks at.
- Every day at a time, or once a week on a day of the week.
- A metrics window: previous complete day, trailing complete days, previous complete week, or month to latest complete day.
A rule can also be triggered by an event rather than a clock — when daily metrics settle, when a group run completes, or on a data-quality event.
Condition
Pick a metric, a comparison, and something to compare against.
| Metric | Reads |
|---|---|
| Platform ROAS | What the platform reports |
| Contribution ROAS | After fees, COGS and shipping |
| Profit-adjusted ROAS | True ROAS — break-even 0 |
| ACOS | Ad spend as a share of ad revenue |
| Attributed GMV | Ad-attributed revenue |
| Profit after ads | What is left once the ad cost is paid |
| Current target ROAS / Current daily budget | The setting as it stands right now |
Behind Show system health metrics sit two more: last successful group runs and consecutive failed group runs — for a rule that should react to the pipeline rather than to performance.
Measure the metric across the whole group or each campaign separately. Comparisons are the usual six (at least, greater than, at most, less than, equals, changed by — absolute or percent), plus three that ask about availability rather than value: is stale, is unavailable, became available.
The threshold is either a fixed number you type, or a share of a setting — for example, 80% of the current target ROAS. Prefer the share: it tracks the setting, so the rule keeps working after you change it.
Several conditions combine with and (every one must be true) or or (any one is enough).
Action
What the rule changes when the condition holds.
| Target | Notes |
|---|---|
| Group Daily Budget / Group Monthly Budget | The group's own limits |
| Group constraint target | The value of the group's constraint |
| Group status | Pause or resume the group |
| Campaign target ROAS | Product Ads or GMV Max |
| Campaign daily budget / Campaign Daily Budget | The platform budget, or the group's per-campaign ceiling |
| Campaign status | Pause or resume a member |
| Use auto bidding | Hand a campaign back to the platform bidder |
| Send notification | Tell you, and change nothing |
Each action is a set to, increase by, decrease by, change by %, or a formula — and carries a direction guard: allow increase or decrease, increase only, or decrease only. Increase-only is the honest choice for a rule that is meant to scale a winner and should never quietly cut anything.
Campaigns
All eligible campaigns in this group, or a list you pick.
Limits
The part to fill in before you turn a rule on, not after.
| Limit | Range | What it is for |
|---|---|---|
| Cooldown | 0–168 hours | Wait after a change before this rule may act again |
| Consecutive matches | 1–14 runs | How many runs in a row must match before anything changes — higher values ignore one-off bad days |
| Changes a day | 0–100 (0 = no cap) | The most changes this rule may make in a day |
| Campaigns a run | up to 100 | The most campaigns one evaluation may touch |
A rule with no cooldown, no daily cap and no consecutive requirement acts on the first match, every time. The panel flags that before you save.
Before you turn it on
The Result section runs the rule against history and against right now:
- "{n} of {total} days would have acted" — the replay. A rule that would have fired on 28 of 30 days is not a guardrail, it is a ratchet.
- "If it runs now" — what it would do against the current numbers, or nothing would change right now.
Save it paused and watch the replay for a few days before activating. You can activate or pause a rule at any time.