Google Ads anomaly detection is the automated monitoring of account performance for statistically significant deviations from baseline. Rather than waiting for a client call or a month-end report, anomaly detection flags issues the morning after they appear. The signal types that matter most for agencies: CPA spikes (today's cost per conversion more than 20% above the 14-day rolling average), CTR drops of 15% or more versus the prior period (often the first visible symptom of ad disapprovals or Quality Score changes), impression share drops of 10% or more (signals a budget cap hit or an increase in auction competition), spend pace anomalies (account on track to overshoot or undershoot monthly budget by 15% or more), and conversion rate drops of 20% or more per campaign (signals landing page issues or audience shift). Each signal has a threshold. Below the threshold, normal variance. Above it, an alert. The difference between catching a CPA spike on day two versus day fourteen is typically several thousand dollars.
What "Anomaly" Means in a Google Ads Context
An anomaly is not just a bad day. It is a statistically significant deviation from the account's own baseline performance, measured over a defined rolling window.
The contrast with manual monitoring matters here. Reviewing campaigns weekly or monthly catches problems after they have already cost money. Anomaly detection runs nightly and flags issues the next morning, before the client calls and before the issue compounds.
At agency scale, this distinction becomes critical. An account manager handling 15-20 client accounts cannot monitor each one daily without automation. Manual daily review of 15 accounts would require 2-4 hours per day before any actual optimization work happens. That time cost is unsustainable, so most agencies settle for weekly reviews, which creates a recency gap.
The concept of "rolling baseline" is worth understanding. Performance thresholds adjust to each account's actual behavior, not a generic industry benchmark. An ecommerce account with $4 CPCs and a B2B SaaS account with $40 CPCs both have anomalies when their own baseline shifts, not when they cross a generic industry threshold. What matters is the deviation from the account's own pattern.
The Five Signals COREPPC's Anomaly Detection Monitors
| Signal | Threshold | Likely Cause | Urgency |
|---|---|---|---|
| CPA spike | 20%+ above 14-day rolling average | Ad disapproval, competitor bid increase, landing page conversion drop | Same day |
| CTR drop | 15%+ versus prior 14-day period | Ad disapproval, Quality Score change, new competitor ads | Same day |
| Impression share drop | 10%+ versus prior period | Budget cap hit, competitor bid increase, eligibility change | Within 24 hours |
| Spend pace anomaly | 15%+ off monthly budget target | Overpacing from high-traffic day, underpacing from account suspension | Within 24 hours |
| Conversion rate drop | 20%+ per campaign versus prior period | Landing page error, audience shift, offer change | Same day |
CPA Spike (20%+ Above 14-Day Rolling Average)
A CPA moving from $40 to $44 (10%) on a single day is noise. A CPA moving from $40 to $52 (30%) is a signal. The 20% threshold separates normal variance from issues that require investigation.
When a CPA spike triggers, the likely causes are: a Smart Bidding algorithm update that temporarily destabilizes bids, a competitor entering the auction with higher bids (which raises CPCs without changing conversion rate), a landing page conversion rate drop (ad clicks at the same rate, fewer convert), or an ad disapproval that reduces coverage on high-performing keywords, raising the effective CPA on remaining traffic.
Each cause has a different fix. Checking them in order, starting with ad disapprovals (visible in the Ads section) and then landing page conversion rate, narrows the cause quickly. See the Google Ads audit tool for a structured diagnostic across the full account.
CTR Drop (15%+ Versus Prior Period)
CTR drops are often the first visible sign of ad disapprovals. When Google disapproves an ad, serving pauses on affected keywords, but the notification is not always prominent. A CTR drop that appears without a corresponding spend change often means ads are serving on fewer keywords than expected.
Beyond disapprovals, CTR drops signal Quality Score changes, new competitor ads pushing organic positions down, and seasonal shifts in search query composition. The distinction matters: a disapproval-driven CTR drop recovers when the ad is fixed. A competition-driven CTR drop requires a different response, typically ad copy refinement or bid adjustment to maintain position.
Catching a 15% CTR drop on day 2 rather than day 8 means 6 fewer days of degraded impression share before anyone acts.
Impression Share Drop (10%+ Versus Prior Period)
Impression share loss is the most undermonitored Google Ads anomaly because it does not appear in the standard metrics view. An account losing 30% of eligible impressions can still show a stable cost per conversion on the impressions it does receive. Nothing in the Campaigns view signals the loss. The issue only surfaces when the "Search Lost IS (Budget)" or "Search Lost IS (Rank)" columns are enabled. COREPPC's anomaly detection flags impression share drops of 10% or more versus the prior rolling period. The most common cause is a budget cap hit: the campaign exhausts its daily budget and exits eligible auctions for the rest of the day. The second most common is a competitor entering the auction with higher bids, which raises the auction threshold and reduces eligibility for lower-bid keywords.
See impression share loss due to budget for how to address budget-constrained impression share problems specifically.
Spend Pace Anomaly (15%+ Off Monthly Target)
A 15% overage on a $20,000/month account is $3,000 in unplanned spend. A 15% underpace means $3,000 in committed budget went unspent, underdelivering the client.
Both are problems. Overpacing is more common: high-traffic days trigger Google's 2x daily budget allowance, and without a monthly cap, campaigns spend into the next week's budget. Underpacing happens when accounts get suspended, when ads face broad disapprovals, or when auction competition drops unexpectedly.
The spend pace anomaly catches both, projected from daily pace. Catching a pacing problem on day 5 of the month allows a correction. Catching it on day 28 means 3 weeks of misaligned delivery.
Conversion Rate Drop (20%+ Per Campaign)
Conversion rate drops are distinct from CPA spikes. A CPA spike can be caused by higher CPCs with a stable conversion rate: more expensive clicks, same close rate. A conversion rate drop suggests the issue is on the landing page or in the audience composition, not in the auction.
When conversion rate drops 20% or more per campaign, the likely causes are: a landing page error (form broken, page returning a 404), an audience shift (the campaign is now reaching different users than its baseline), or an offer change on the landing page (price increase, product sold out, CTA removed). All three require investigation outside the Google Ads interface itself.
Why Manual Monitoring Fails at Agency Scale
Manual Google Ads monitoring has a compounding recency problem at agency scale. An account manager reviewing 15 accounts weekly spends approximately 5 hours on review before any optimization work begins. A weekly review cadence means a CPA spike that started on day 2 of the week is not caught until day 7 or day 8, after the issue has run for 6-7 days. On a $500/day account, that is $3,000-$3,500 in spend at degraded performance before anyone noticed. The problem is not the manager's skill or attention. It is the economics of weekly review at scale. Anomaly detection solves the recency gap by running every night and surfacing deviations the next morning, before client calls and before the issue compounds further.
Four specific failure modes show up consistently in manual-monitoring agency setups:
-
1.
Time cost
Reviewing 15 accounts at 20 minutes per account is 5 hours per week, before any actual optimization work. In practice, this means reviews get shorter or less frequent.
-
2.
Recency gap
Weekly review catches a CPA spike 4-7 days after it started. Seven days at degraded performance on a $500/day account is $3,500 spent before anyone noticed.
-
3.
Threshold inconsistency
Manual reviewers apply different internal thresholds depending on the day, the account, and the reviewer. Automated thresholds are consistent. The same 20% CPA deviation triggers an alert on every account, every night, regardless of who is on the team.
-
4.
Coverage gaps
Manual monitoring is typically more rigorous for large-spend accounts. Small-budget client accounts get less attention and accumulate problems faster. Anomaly detection covers every account in the portfolio equally, every night.
COREPPC's anomaly detection runs nightly. Agencies see alerts the next morning, before the client calls.
Set up anomaly monitoring for your agency portfolio. Alerts the next morning, before any client calls.
Set Up Anomaly MonitoringHow to Set Up Anomaly Detection for Google Ads Accounts
Two options exist: native Google Ads alerts and third-party monitoring tools.
Native Google Ads alerts: Go to Tools and Settings > Alerts. Google allows setting alerts for budget, conversions, and CPC. The limitations are meaningful: alerts require manual configuration per account, they use absolute thresholds rather than rolling baselines (so a campaign that normally runs at $50 CPA triggers the same alert as one that normally runs at $15 CPA), and they do not cover impression share or conversion rate drops. For agencies with multiple client accounts, this means logging in to each account and configuring alerts separately.
Third-party monitoring, including COREPPC: Third-party tools add rolling baseline calculations (14-day), multi-account coverage from one dashboard, and alert types beyond what native Google supports: impression share drops, spend pacing anomalies, conversion rate changes per campaign.
The COREPPC setup takes three steps:
One-time setup. Covers all linked sub-accounts automatically. No data exports required.
Email alert, in-dashboard flag, or both. Threshold logic is fixed and consistent across all accounts in the portfolio.
Alerts surface before the business day begins, before client communication starts. Any account that breached a threshold gets flagged for investigation.
This is a monitoring workflow, not a replacement for structural auditing. The audit finds configuration problems. Anomaly detection catches ongoing deviations. Both are part of the same agency process. For how this fits into the full reporting cycle, see Google Ads agency reporting. For identifying the structural problems anomaly detection surfaces, use the guide to detecting budget waste before it compounds.
Set Up Anomaly Monitoring
Nightly monitoring across all your client accounts. Alerts before the client calls.
Set Up Anomaly MonitoringFrequently Asked Questions
Google Ads anomaly detection is the automated monitoring of account performance for statistically significant deviations from a rolling baseline. It flags metrics that have shifted beyond defined thresholds: CPA spikes of 20%+ above the 14-day average, CTR drops of 15%+ versus prior period, impression share drops of 10%+, spend pace anomalies of 15%+ off monthly target, and conversion rate drops of 20%+ per campaign. These alerts fire the morning after the anomaly appears, before it compounds.
Google's native alerts (Tools and Settings > Alerts) allow setting absolute thresholds for budget, conversions, and CPC. These are manually configured per account and use fixed values rather than rolling baselines. They do not cover impression share or conversion rate drops, and they do not scale to multi-account monitoring without manual setup on each account individually.
A CTR drop of 15% or more is often the first symptom of ad disapprovals, which can pause ad serving on affected keywords without a prominent notification. CTR drops also signal Quality Score changes, new competitive ads pushing positions down, and seasonal shifts in search query composition. Catching a CTR drop early preserves impression share before it cascades into CPA degradation.
Automated anomaly detection should run nightly and surface alerts the following morning. Manual review for agencies managing 10+ accounts: daily alert review (5-10 minutes), deeper investigation for any account that triggered an alert. Weekly full review for accounts with no active alerts. Monthly performance review for all accounts regardless of alert status.
First, check if CPA is 20%+ above the 14-day rolling average, or if it is within normal variance. If it is a real spike: check whether any ads were disapproved (which reduces coverage and raises effective CPA), check for competitor bidding changes in Auction Insights, check whether a Smart Bidding algorithm update coincided with the spike date, and check landing page conversion rate independently (separate from Google Ads reported conversions).
COREPPC monitors connected Google Ads accounts nightly, measuring five signals against a 14-day rolling baseline: CPA, CTR, impression share, spend pace, and conversion rate per campaign. Accounts that breach any threshold get flagged in the agency dashboard the following morning. The alerts appear before the start of the business day, before client communication begins. The threshold logic is fixed: a CPA 20%+ above baseline triggers regardless of account size or campaign type.
By Dror Aharon | April 28, 2026