Facebook ads automated rules on ROI: stop-loss and scale

Meta’s automated rules judge a campaign by what Meta can see: spend, cost per result, CPM, frequency. A buying team is paid on what the advertiser confirms later, such as a deposit that only the tracker or the advertiser’s statistics know about. So a cabinet rule on CPI or cost per result can pause a campaign that earns and keep one that burns. The fix is to run the same stop-loss and scaling logic on tracker profit and ROI, with a minimum spend before any decision, a limit on every action, and a backtest and a review before the rule acts alone.

Why cabinet rules pause the wrong campaigns

Meta describes automated rules as checks of campaigns, ad sets and ads that then update them or notify you. A rule has a condition, an action and the objects it applies to. The conditions Meta lists, as of November 2026, are Meta’s own measurements: cost per result, per purchase, per lead, per registration or per app install, CPM, impressions, reach, frequency, results, budget and bid values, and timing fields. Payout is not among them. The advertiser pays when a deposit is confirmed, and that fact lives in the tracker.

An example with invented figures. Two campaigns spent 60 each today.

	Spend	Installs	CPI in cabinet	Deposits	Revenue	Profit	ROI
Campaign A	60	12	5.00	3	120	60	100%
Campaign B	60	30	2.00	0	0	-60	-100%

A cabinet rule that pauses anything with CPI above 4 stops Campaign A, the winner, and leaves Campaign B running. Cheap installs that never deposit look perfect in the cabinet. The tracker sees what the cabinet cannot.

There are other reasons cabinet rules hurt a buying team:

• Attribution and delay. Meta’s numbers can lag and differ from the tracker’s, as the profit per media buyer guide explains. A rule that fires on early numbers acts on incomplete data.
• No view of caps. A campaign can be profitable and still pointless to scale because the advertiser’s cap is nearly full. A cabinet rule does not know. Read the caps guide for how to count that.
• Hard limits. One Meta rule applies to one level only, so campaigns and ad sets need separate rules. An ad account holds up to 250 rules, active and inactive together, and each condition can be used once per rule. Meta also says that rules can’t run on ads about social issues, elections or politics, and can’t pause reservation campaigns.
• Several conditions mean all of them. If you add more than one condition, the object has to meet every one. For “either this or that” you need two rules.

Meta’s own rules still have a place: a notification when a lifetime budget is nearly spent, a schedule that turns a campaign on and off, a frequency cap. Use them for things that really are about the cabinet.

Rules on tracker profit and ROI

A rule on profit needs the same two sides as a report: spend from the ad account and revenue from the tracker, joined on the same campaign. The numbers you want in a condition:

• Spend, to avoid deciding on too little data.
• Revenue and profit, from confirmed conversions, by period.
• ROI, profit divided by spend.
• EPC, revenue per click, for early traffic when deposits have not arrived yet.
• Cap left, so a winner is not scaled into a full cap.
• Your own metric, if the team has its own definition, such as ROI after the agency fee.

Pick the time window with the delay in mind. Deposits arrive minutes, hours or days after the click. A rule that asks “ROI below zero today” at 10 in the morning will condemn every campaign. Better to say “spent at least X over the last 24 hours and ROI below Y”, or to judge on yesterday and the day before. Also note the freshness of each input: tracker conversions land at some interval, and spend lands at another. Check both before trusting a rule on short windows.

Choose the level on purpose. Tracker data is clean at the campaign level, because the link from click to campaign is the campaign parameter. Ad sets and ads depend on how well your sub parameters carry that detail. Start at the campaign.

Stop-loss rules

A stop-loss rule protects money, so it should be simple and boring. A good one has four parts:

1. A floor on spend. “Spent more than X in the window.” Zero deposits after a small spend means nothing; after enough spend it does.
2. A threshold on result. “ROI below Y” or “no confirmed revenue”.
3. An action with a way back. Pause for a set time, not forever, and notify the buyer. A paused campaign that wakes up by itself can be re-judged on fresh data.
4. A cap on frequency of action. One pause per campaign per day is enough.

Two stop-loss levels work well together. A fast one catches a disaster: spend of 2X with zero revenue, pause. A slow one catches a slow leak: ROI below the team’s minimum over three days, notify the buyer and wait for their decision. The fast rule can act alone. The slow one is a conversation.

Avoid stop-loss rules that look at one hour. Variance in a single hour is large, and the rule will cut campaigns that would have recovered.

Scaling and duplicating winners

Scaling is the mirror of stop-loss, and the same discipline applies: enough spend, a result threshold, a limit on action.

• Raise the budget in small steps. Meta’s budget rules can increase or decrease a budget by a percentage or a fixed amount, and set a maximum or minimum. Use both: a step that is small enough not to disturb the campaign and a ceiling that is yours, not the account’s.
• Check the cap before scaling. More budget on a nearly full cap is waste. Put “cap left” into the condition.
• Do not scale on one good day. Require ROI above the threshold over a window with real spend.
• Duplicate instead of raising the budget when you do not want to touch a campaign that works. A copy takes its own budget and can be paused without touching the original.
• Limit the copies. A set number of duplicates per day, copies starting paused, and a review before they go live, so one lucky campaign does not become twenty.

Scaling mistakes cost more than stop-loss mistakes, because a bad scale spends faster. For this reason many teams let stop-loss act alone and put scaling behind a review.

Backtest and review before the rule acts alone

A rule is code that spends your money. Test it before it has the keys.

1. Backtest on past days. Run the rule over the last few weeks and list what it would have done. Count the campaigns it would have paused that later made money. If it paused winners, the threshold or the window is wrong.
2. Dry run. Let it run live but without acting, and compare its list with what the buyers did by hand.
3. Review mode. Let the rule act only after a person approves each action. When approvals are almost always yes, let it act alone.
4. A log of every run. What the rule saw, which campaigns matched, what it did. A rule you cannot audit is a rule you will switch off after the first bad day.
5. A weekly look. Compare the rule’s decisions with outcomes. Thresholds drift as the market, the offer and the creative change.

On TikTok, remember the difference in tools. TikTok’s own help page describes rules that send notifications, pause underperforming ads and increase budgets or bids, with the target set at campaign, ad group or ad. If you drive TikTok from a third-party tool, check which of those actions the tool can use.

Auto-rules in DarkCore CRM, honestly

Auto-rules in DarkCore CRM read what the tracker says a campaign earns next to what the ad account says it costs. A condition can use spend, CPI, CPM and CTR from the cabinet, revenue, profit, ROI, EPC and cap left from the CRM, events, and your own metrics. Tracker metrics work on campaigns. A rule can look at today, yesterday or the last 3, 6 or 24 hours, and runs every 5, 15 or 30 minutes, hourly or once a day. Inputs come from Meta and TikTok ad accounts and from Keitaro and DarkCore PWA trackers.

A rule can act right away, only notify, or wait for a person to review each action. On Meta and TikTok it can pause for a set time, activate or notify. On Meta only, it can also set or shift a budget within a floor and a ceiling and duplicate a winner a set number of times a day, and the copies start paused. Before a rule goes live you can backtest it on past days and run it dry to preview the matches, and every run is logged with the campaigns it matched and the action it took.

The limits are plain. On TikTok rules only pause, activate, notify and wait for review: budget changes and duplicates are Meta only. Tracker metrics work on campaigns, not on ad sets or ads. The site does not state how fresh each metric is at the moment a rule runs, so ask for that in the demo, using your own tracker as the example. The site shows no prices, and a demo is booked in Telegram.

FAQ

Can Meta’s automated rules use tracker ROI?

Not by themselves. The conditions in Meta’s help center are Meta’s own measurements, such as cost per result, CPM, frequency, results and budget values. To act on tracker profit or ROI you need a tool that reads the tracker and the cabinet together and then pauses or changes the campaign.

How much spend should a stop-loss rule wait for?

Enough that one event cannot decide. One way is to set the floor as a multiple of the expected cost of one confirmed result on the offer, and to adjust it after a backtest. Your payout, your cost per deposit and your conversion delay set the right number, so there is no universal value.

Is it safe to let a rule scale budgets alone?

Only after a backtest, a dry run and a period in review mode, and only with a ceiling on the budget, a limit on steps per day and a check of the cap. Many teams let stop-loss rules act alone and keep scaling behind a human review.