Alerts and monitoring

Updated November 12, 2026

What AdsGuru alerts on

Four conditions interrupt you immediately; everything else waits for the daily note:

• Tracking breaks: conversion volume collapsing while clicks hold, the signature of a measurement failure rather than a demand failure.
• Runaway spend: pacing far above plan, the day it starts, before a bad week becomes a bad month.
• Disapprovals and policy flags: ads or assets rejected on either network, especially clusters that suggest an account-level issue.
• Anomalies: CPM, CPA or delivery swings beyond the account's normal volatility that do not match any known cause.

The bar for interrupting is deliberately high. An alert channel that cries wolf gets muted, and then the real wolf walks in unannounced. And everything on this page concerns watching rather than acting: an alert can arrive with a proposed response attached, and the response still goes through the same approval gate as any other change. Monitoring that quietly acts on its own findings is how accounts get thrashed at 3 a.m.

Tracking breaks: the costliest silent failure

A tracking break means conversions stop being recorded while the ads keep running, and its signature is specific: clicks and spend hold steady while reported conversions collapse. Real demand rarely vanishes overnight with traffic intact; measurement does it all the time, after a site update, a consent banner change, a tag manager edit, a platform integration hiccup.

It is the costliest failure because everything downstream reacts to the false zero: smart bidding sees no conversions and flails, budget logic wants to cut the 'failing' campaigns, and a human looking at dashboards panics in the wrong direction. The correct first response to a conversion cliff is always to suspect measurement before demand.

When AdsGuru detects the signature, the alert says exactly that: conversions fell off a cliff at this hour while clicks held, here is the likely layer (pixel, CAPI, conversion action), verify the tag or recent site changes. During a suspected break, optimization routines hold rather than reacting to data known to be broken. The audit's tracking checks are the preventive version of this alert.

Runaway spend

Runaway spend means the account is pacing far above plan, and it alerts the day the run rate breaks the threshold rather than when the money is gone. Causes, roughly in order of frequency: budget changes nobody meant (including auto-applied recommendations), a bid strategy recalibrating after an edit, a new campaign set to a daily budget someone meant as monthly, and platform pacing quirks where a daily budget legitimately spends up to about double on a given day while the month balances.

That last cause is why the alert reads pacing, the projected month, instead of single-day spend: one heavy Tuesday is normal platform behavior, while three heavy days that project the month 40% over plan are an emergency. The alert names the campaigns driving the overage and proposes the trim; budget automation carries the correction after your approval.

The reverse condition, underdelivery, is monitored too, but flags in the daily note rather than interrupting: money not spent is an opportunity problem, not a bleeding one.

Disapprovals and policy flags

Disapprovals alert immediately because they compound quietly: a disapproved ad simply stops serving, spend shifts to whatever remains eligible, and an account can drift into serving its worst creative with nobody deciding anything. A cluster of disapprovals, or a policy flag at the account level, can precede worse: limited delivery states or account suspension.

The alert carries the platform's stated reason, which asset or ad is affected, what stopped serving, and where spend went instead. Responses split by case: a false-positive disapproval goes to appeal in the platform; a legitimate one needs the creative or landing page fixed; a repeated pattern, like a claim that keeps tripping policy, needs the claim retired from the creative rotation.

AdsGuru cannot file appeals for you, platforms require the account owner, but the alert includes exactly where to click and what the policy objection was, which removes most of the friction.

Anomalies: separating weird from normal

An anomaly alert fires when a metric breaks the account's own volatility band and no known cause explains it. The baseline is the account's history: its weekday rhythm, its seasonal shape, its normal CPM wobble, learned over the first 7 to 14 days and refined after. This is why anomaly detection matures after the first two weeks: without a baseline, everything or nothing looks weird, and both are useless.

The known-cause check is what keeps anomaly alerts rare: a CPM spike during a holiday auction, a CTR drop explained by creative fatigue already being tracked, or a delta following a change you approved yesterday are all explained movements, reported with their causes in the daily note instead of alerting. What alerts is the residue: movements that match nothing.

An anomaly alert states the metric, the size of the break against the baseline, what was ruled out, and the current best hypothesis, plus what evidence would settle it. Sometimes the honest answer is 'unknown, watching'; pretending certainty would poison the channel.

The baseline also updates continuously: a promotion week you run every month becomes part of normal, and the system stops treating it as news. What never becomes normal is a pattern that keeps costing money without explanation.

Why routine optimization is batched

Everything that is not an emergency waits for the daily batch, on purpose. Ad platforms penalize thrash: frequent edits reset Meta learning phases and keep Google bid strategies perpetually recalibrating, so a system reacting to every hourly wobble makes the account measurably worse while looking vigilant.

Batching also respects your attention: one note a day you actually read beats fourteen pings you learn to swipe away. The division of labor is strict: emergencies interrupt, optimization batches, and nothing in either lane applies without approval or an autopilot grant.

Alert reference

Monitor an account manually

The manual monitoring routine AdsGuru automates, for reference or for running alongside it:

1. Daily, two minutes: yesterday's spend and conversions per platform against plan; any ad or asset disapprovals (Google: Ads status column; Meta: Account Quality page).
2. Daily: eyeball conversions against clicks; a conversion cliff with steady clicks means check tracking before touching campaigns.
3. Weekly: pacing math (month-to-date spend versus plan), frequency on main Meta ad sets, impression share trends on key Google campaigns.
4. Weekly: skim change history on both platforms for changes you do not recognize, including auto-applied ones.
5. Monthly: full audit pass: tracking, waste, settings, structure.

The discipline is doing it every day, including the boring ones. That consistency is precisely what software is better at than people, which is the honest case for automating the watch and keeping the decisions. If you do run the checklist yourself, keep a dated log of what you checked and saw: monitoring without a record degrades into vibes within a month, and the record is what lets you tell a trend from a mood.

Common questions

Related