The Media Buyer's Guide to Automating Meta Ads Rules

You went to bed at midnight. ROAS was 3.2x, CPA was within target, spend was pacing well. You slept six hours. You woke up to $2,400 burned on a campaign that tanked at 2 AM because a creative hit frequency saturation and Meta's algorithm decided to chase reach over conversions. Your daily budget was gone by 8 AM.

This is the "I'll check it in the morning" problem. And every media buyer who's been at this long enough has a version of this story. The specifics change -- sometimes it's a CPA spike, sometimes it's a broken landing page, sometimes it's an audience that suddenly stops converting -- but the pattern is always the same: campaigns go sideways between checks, and by the time you notice, money has been wasted.

The solution is automation. But not the kind that Meta offers natively.

---

Why Meta's Native Rules Fall Short

Meta Ads Manager does have an automated rules feature. It's... basic. You can create rules based on a handful of metrics, with limited logic (one condition at a time or simple AND), and the actions are restricted to pause, send notification, or adjust budget.

Here's what's missing:

The fundamental difference: Meta's rules are a basic safety net. AdRow's rules engine is a programmable decision layer that operates on your entire ad account ecosystem simultaneously.

---

Anatomy of a Good Automation Rule

Every rule in AdRow has four components: conditions, logic, filters, and actions. Understanding how they interact is the difference between a rule that saves you money and a rule that causes chaos.

Conditions

A condition is a statement about a metric:

metric: "cpa"
operator: "gt"
value: 45
time_range: "today"

This translates to: "CPA today is greater than $45."

The available metrics break down into two categories:

Raw metrics (collected directly from Meta API):

Derived metrics (calculated by the engine):

That's 25 metrics you can use as condition inputs. Each condition also has a time range -- the window over which the metric is aggregated: today, yesterday, last_3d, last_7d, last_14d, or last_30d.

Operators determine how the metric is compared to your threshold:

AND/OR Logic

Rules support AND and OR condition logic:

• AND (default): All conditions must be true for the rule to trigger. Use this for most rules. Example: CPA > $40 AND spend > $50 -- both must be true.
• OR: Any condition being true triggers the rule. Use this for catch-all safety nets. Example: CPA > $80 OR ROAS < 0.5 -- either means something is very wrong.

Entity Filters

Before conditions are even evaluated, you can filter which entities the rule applies to:

• Name contains: Only evaluate campaigns/ad sets/ads whose name includes a specific string. Useful for targeting specific naming convention segments (e.g., only campaigns with "US_" in the name).
• Name not contains: Exclude entities by name pattern. Keep your evergreen campaigns safe from aggressive rules.
• Status filter: Only evaluate entities with a specific status (ACTIVE, PAUSED, etc.).

This is important. Without filters, a rule applies to everything in the selected ad accounts. Filters give you surgical precision.

---

Action Types

When a rule's conditions match, one or more actions execute. Here's what each does.

Pause (The Kill Switch)

The most common action. Pauses the matching campaign, ad set, or ad via the Meta API. Immediate. No gradual wind-down -- the entity stops delivering.

Use for: CPA guards, frequency caps, zero-conversion killers, budget blowout protection.

Activate (The Resurrection)

Re-activates a paused entity. This is the counterpart to pause -- used in rules that monitor paused entities for historical performance and bring them back when metrics indicate potential.

Use for: Re-activating top performers that were paused during a down period.

Increase Budget % (The Growth Lever)

Increases the entity's daily budget by a specified percentage. This is the scaling action.

action: "increase_budget_pct"
params: { percentage: 15 }

This increases the budget by 15%. If the current budget is $100/day, it becomes $115/day.

Critically, this action works in concert with safeguards (see below). You should never use budget increase without setting budget_change_limit_pct and budget_daily_cap.

Decrease Budget % (The De-escalation)

The defensive counterpart. Reduces budget by a percentage when performance degrades but isn't bad enough to justify a full pause.

action: "decrease_budget_pct"
params: { percentage: 20 }

Use for: Gradual wind-down of campaigns that are marginally profitable but trending down.

Notify Only (The Observer)

Sends a notification (Telegram, in-app, or both) without taking any action on the entity. The rule evaluates and matches, but the only output is a message.

Use for: Monitoring unusual patterns, tracking metric thresholds, getting alerted about conditions that require human judgment before action.

---

5 Rule Templates

These are concrete, battle-tested configurations. Copy them, adjust the thresholds to your margins and scale, and deploy.

1. Kill Losers (CPA Guard)

The most important rule in any account. Catches campaigns that are spending without converting at your target CPA.

Rule: CPA Guard - Kill Losers
Entity level: Ad Set
Condition logic: AND

Conditions:
  [1] metric: cpa      | operator: gt  | value: 45   | time_range: today
  [2] metric: spend     | operator: gt  | value: 50   | time_range: today

Actions:
  [1] type: pause

Safeguards:
  cooldown_minutes: 360         (6 hours)
  max_executions_per_day: 3
  notify_on_execution: true
  notify_channels: { telegram: true, in_app: true }

Entity filter:
  status_in: ["ACTIVE"]

Why both conditions? Condition [1] alone would pause ad sets that spent $2 on 0 conversions (CPA = infinity). Condition [2] ensures the ad set has spent enough to be statistically meaningful before the rule triggers. Adjust $50 based on your product price -- for a $200 product, you might set this at $150 or even $200 to give the algo room.

Threshold math: If your target CPA is $25, setting the kill threshold at $45 (1.8x) gives campaigns a reasonable learning phase. Setting it at $50 (2x target) is more conservative. Setting it at $30 (1.2x) is aggressive and may kill campaigns still in the learning phase.

2. Scale Winners (ROAS Autopilot)

Automatically increases budget on campaigns that are crushing it.

Rule: ROAS Autopilot - Scale Winners
Entity level: Campaign
Condition logic: AND

Conditions:
  [1] metric: roas     | operator: gt  | value: 3.0  | time_range: last_3d
  [2] metric: spend     | operator: gt  | value: 100  | time_range: last_3d
  [3] metric: purchases | operator: gte | value: 5    | time_range: last_3d

Actions:
  [1] type: increase_budget_pct
      params: { percentage: 15 }

Safeguards:
  cooldown_minutes: 360         (6 hours)
  max_executions_per_day: 3
  budget_change_limit_pct: 30   (max 30% budget increase per day)
  budget_daily_cap: 500         (absolute max daily budget: $500)
  notify_on_execution: true

Three conditions, not one. Condition [1] checks ROAS. But ROAS alone is misleading on small spend -- $10 spend with $50 revenue is 5x ROAS but statistically meaningless. Condition [2] ensures sufficient spend. Condition [3] ensures sufficient conversion volume. All three must be true (AND logic).

Budget safeguards are non-negotiable. The rule increases budget by 15% per execution. With max 3 executions/day and 6-hour cooldown, the maximum daily increase is 15% compounded 3 times = ~52% in theory, but the budget_change_limit_pct: 30 caps total daily increase at 30%. And budget_daily_cap: 500 sets a hard ceiling. These safeguards prevent a rule from scaling a $50/day campaign to $5,000/day overnight.

Why last_3d instead of today? Today's data is noisy, especially early in the day. A 3-day window smooths out daily variance and gives a more reliable signal. For products with longer conversion windows, consider last_7d.

3. Frequency Cap

Catches ads that are being shown to the same people too many times, killing creative effectiveness.

Rule: Frequency Cap
Entity level: Ad
Condition logic: AND

Conditions:
  [1] metric: frequency   | operator: gt  | value: 3.0  | time_range: last_3d
  [2] metric: impressions  | operator: gt  | value: 5000 | time_range: last_3d

Actions:
  [1] type: pause

Safeguards:
  cooldown_minutes: 360
  max_executions_per_day: 3
  notify_on_execution: true

Entity filter:
  status_in: ["ACTIVE"]

At the ad level, not campaign. Frequency issues are creative-specific. A campaign might have 4 ads where one has hit frequency saturation while the others are fine. Pausing at the ad level preserves the healthy ads and the campaign's overall learning.

Condition [2] filters out low-volume ads. An ad with 100 impressions and frequency 3.5 is probably fine -- it just hasn't had enough volume for the frequency to normalize. 5,000 impressions is enough to be confident the frequency is a real problem.

Threshold of 3.0 over 3 days is moderate. For direct response, some media buyers use 2.5. For brand awareness, you might go up to 5.0. The right number depends on your audience size and creative rotation strategy.

4. Zero Conversion Alert

Catches campaigns that are spending with absolutely zero results. This isn't about bad CPA -- it's about detecting campaigns that are fundamentally broken (wrong pixel, broken landing page, wrong optimization event).

Rule: Zero Conversion Alert
Entity level: Ad Set
Condition logic: AND

Conditions:
  [1] metric: conversions | operator: eq | value: 0   | time_range: today
  [2] metric: spend        | operator: gt | value: 30  | time_range: today

Actions:
  [1] type: notify_only

Safeguards:
  cooldown_minutes: 360
  max_executions_per_day: 3
  notify_channels: { telegram: true, in_app: true }

Entity filter:
  status_in: ["ACTIVE"]

Notify, don't pause. Zero conversions today could mean the campaign is in learning phase, or the conversion window hasn't closed, or it's 10 AM and conversions typically come in the afternoon. This rule alerts you so you can investigate, rather than automatically killing campaigns that might be fine.

$30 threshold is adjustable. For low-ticket products, $30 of zero-conversion spend is already a red flag. For high-ticket ($500+), you might set this at $100-200. The point is to catch anomalies before they become expensive.

Telegram notification is key here. The whole point of this rule is speed of notification. You want to know about zero-conversion spend as soon as the rule evaluates -- which happens every 15 minutes. Within 15 minutes of hitting $30 spend with zero conversions, you get a Telegram message with the campaign name, spend, and metrics. That's fast enough to investigate and fix the issue before the daily budget is exhausted.

5. Re-activate Top Performers

The rule most people forget to set up. Scans paused campaigns for recent historical performance that suggests they deserve another chance.

Rule: Re-activate Top Performers
Entity level: Campaign
Condition logic: AND

Conditions:
  [1] metric: roas       | operator: gt  | value: 2.0  | time_range: last_14d
  [2] metric: purchases   | operator: gte | value: 10   | time_range: last_14d

Actions:
  [1] type: activate

Safeguards:
  cooldown_minutes: 360
  max_executions_per_day: 3
  notify_on_execution: true

Entity filter:
  status_in: ["PAUSED"]
  name_not_contains: "[DEAD]"

Targets PAUSED campaigns. The entity filter ensures this rule only looks at campaigns that are currently paused. If a campaign was paused (manually or by another rule) but its 14-day performance shows 2x+ ROAS with 10+ purchases, it gets reactivated.

14-day lookback is intentional. A campaign that did well over 14 days has a real signal, not a one-day fluke. This is long enough to smooth out variance but short enough to be relevant.

The name_not_contains: "[DEAD]" filter is a convention trick. If you manually pause a campaign and rename it to include "[DEAD]", the re-activation rule won't touch it. This lets you permanently pause campaigns without them being resurrected by automation. It's a simple naming convention that acts as a human override.

Use with the CPA Guard. These two rules create a loop: the CPA Guard pauses losers, the Re-activator checks if paused campaigns have strong historical data, and re-enables them. Campaigns can cycle through pause/re-activate until they either consistently perform or get manually tagged as [DEAD].

---

Safeguards: Why They Matter More Than Rules

Here's a truth nobody tells you about automation: the rules themselves are easy. Writing "if CPA > $40, pause" takes 30 seconds. What's hard is preventing automation from doing something catastrophic.

Every safeguard in AdRow exists because someone, somewhere, learned an expensive lesson.

Cooldown (default: 6 hours)

After a rule acts on an entity, it won't act on the same entity again for the cooldown period. This prevents rapid-fire actions.

Without cooldown, a scale rule could increase budget, the next evaluation cycle (15 minutes later) sees the budget increase reflected in spend but not yet in conversions, and increases budget again. And again. And again. In 2 hours, your $100/day campaign could be at $800/day.

6 hours is the default because it gives Meta's algorithm time to re-optimize after a budget change before the next evaluation.

Max Executions Per Day (default: 3)

Caps how many times a rule can execute actions per calendar day, across all entities. Even with cooldowns, a rule matching 50 entities could execute 50 actions in one cycle. Max executions caps the total.

3 per day is conservative. For monitoring-heavy rules (notify_only), you might increase this. For budget adjustment rules, 3 is appropriate -- you rarely want to change budgets more than 3 times in a day.

Budget Change Limit %

Sets the maximum total budget change per day as a percentage. If budget_change_limit_pct: 30, a campaign's budget can't increase or decrease by more than 30% in a single day, regardless of how many times the rule triggers.

This is your backstop against compounding. A 15% increase executed 3 times (with cooldown allowing it) compounds to ~52%, but the 30% cap catches it.

Budget Daily Cap

An absolute ceiling on budget. If budget_daily_cap: 500, the campaign's daily budget will never exceed $500, period. No matter how many times the scale rule triggers, no matter how good the ROAS is.

This is your "sleep at night" number. Set it to the maximum you're willing to spend on a single campaign per day.

Audit Log

Every rule execution is logged with full detail: which entities matched, what metrics they had, what actions were taken, and the outcome. You can review this in the Dashboard execution history. This isn't just for debugging -- it's for optimization. Review your rule execution logs weekly to calibrate thresholds.

Check execution history in the Reports section to correlate rule actions with performance changes over time.

---

Setting Up Telegram Notifications

Telegram notifications are the fastest way to stay informed about rule executions. Here's the setup.

1Step 1: Link Your Telegram Account

In AdRow, go to Settings > Notifications > Telegram. Click "Link Telegram" to get a unique link token. Open Telegram, find the AdRow bot, and send the token. The link is confirmed instantly.

2Step 2: Configure Notification Preferences

You control which events trigger Telegram messages:

• Rule executed: Get notified when any automation rule takes action (recommended: ON)
• Rule errors: Get notified when a rule fails to execute (recommended: ON)
• Other notification types (spend alerts, daily digests, etc.) are also available

3Step 3: What You Receive

When a rule executes, you get a structured message like this:

Rule "CPA Guard - Kill Losers" -- 3 adsets matched
Actions: paused

- US_LAL1_VideoA
  Spend: $67.50 | ROAS: 0.80x | Conv: 1
- US_Broad_ImageB
  Spend: $52.30 | ROAS: 0.45x | Conv: 0
- DE_Interest_VideoC
  Spend: $58.10 | ROAS: 0.92x | Conv: 1

Each notification includes: rule name, number of matched entities, actions taken, and for each entity: name, key metrics (spend, ROAS, conversions). If more than 20 entities match, the message is truncated with a link to the platform for full details.

Messages are sent in real-time via the Telegram Bot API -- there's no batching delay. The rule evaluates, actions execute, and the notification fires within the same cycle. Given the 15-minute evaluation frequency, worst case you're notified within 15 minutes of a metric threshold being crossed.

---

Putting It All Together: A Rule Stack

Individual rules are useful. A coordinated set of rules is a system. Here's a production-ready rule stack for an e-commerce account:

Rules are evaluated in priority order. Higher priority rules execute first. If the CPA Guard pauses a campaign, the Scale Winners rule won't try to increase its budget in the same cycle.

This stack creates a self-correcting system:

• New campaigns launch (via the Launcher or Ads Manager)
• CPA Guard kills losers early
• Scale Winners amplifies what's working
• Frequency Cap protects creative freshness
• Re-activator gives paused campaigns a second chance
• Zero Conv Alert catches broken setups
• Budget Blowout Guard is the ultimate safety net

You monitor the system through notifications and the Dashboard, intervening only when rules flag something unusual or when you need to make strategic changes (new creatives, new audiences, new geos).

---

Key Takeaways

1. Automate defense first, offense second. Your first rules should be pause rules (CPA Guard, Frequency Cap). These protect your budget. Scale rules come after you've proven the defense layer works. An aggressive scale rule without a CPA Guard is a recipe for burning money fast.

2. Every budget action needs a safeguard. Cooldowns, max executions, budget change limits, and daily caps are not optional. They're what prevent automation from doing in 15 minutes what would take you months to recover from.

3. Thresholds are not permanent. Set initial thresholds based on your current averages, then review and adjust weekly. Your rule execution history is a dataset -- use it. If a rule triggers 50 times a day, the threshold is too tight. If it never triggers, it's too loose or your campaigns are doing well.

4. The 15-minute evaluation cycle is your resolution floor. Rules check every 15 minutes. If your daily budgets are high enough that significant spend can happen in 15 minutes, set tighter thresholds or lower your daily budgets. At $1,000/day, ~$10 can be spent per 15-minute window. At $10,000/day, that's ~$100.

5. Use naming conventions as control mechanisms. The [DEAD] convention for permanently paused campaigns, geo prefixes for geo-specific rules, and test markers for experimental campaigns -- these naming patterns become part of your automation infrastructure through entity filters.

6. Telegram is not optional for real-time operations. Email notifications are fine for daily summaries. For rules that pause or scale campaigns, you want Telegram. The latency difference between "I'll see the email in an hour" and "my phone buzzed 30 seconds ago" is the difference between $50 and $500 of wasted spend.

7. Rules create feedback loops -- design them intentionally. A CPA Guard that pauses and a Re-activator that un-pauses create a cycle. This is by design, not a bug. The cycle stabilizes when a campaign either consistently performs (stays active) or consistently fails (gets manually tagged as [DEAD]). Understand these loops before deploying them.

The goal of automation is not to remove humans from the process. It's to move the human role from "check campaigns every 2 hours and make repetitive decisions" to "design the decision framework, review results, and intervene on edge cases." The rules do the watching. You do the thinking.