Catch performance anomalies before they become disasters
Analyze spend, conversion, and delivery fluctuations against rolling averages to separate normal variance from real problems
Critical issues might be hiding in yesterday's performance data while you focus on monthly reports
This template performs statistical anomaly detection across all campaign metrics to catch problems within 24 hours instead of weeks. It compares day-over-day and week-over-week changes against 14-day rolling averages, using impact weighting to separate normal fluctuations from genuine issues that need immediate attention.
Why this matters
A tracking pixel that breaks on Tuesday won't show up in your weekly report until Monday—after losing six days of conversion data. A campaign silently hitting frequency caps looks fine in spend reports but bleeds performance daily. This template catches sudden conversion drops with steady clicks (tracking issues), impression collapses with unchanged budgets (disapprovals), and gradual 5-day declines that signal structural problems. Each anomaly comes with a likely diagnosis and specific next steps.
How to customize
Adjust the rolling average window based on your typical conversion lag—B2B might need 21-30 days. Add platform-specific metrics like Quality Score changes or Relevance Score drops for deeper diagnosis. Include external calendar events like sales periods or holidays to reduce false positives. Set different sensitivity thresholds by campaign type since brand campaigns naturally have less variance than prospecting. Configure alerts to fire only for campaigns above certain spend thresholds to focus on material impact.
You are auditing campaigns for significant performance fluctuations. Your goals:
* Identify unusual DoD and WoW shifts in spend, clicks, conversions, ROAS, CPA, impressions
* Distinguish normal variance from anomalies
* Diagnose likely causes and recommend next steps
* Prioritise issues by impact and urgency
**Data you have access to**
* Daily spend, clicks, conversions, ROAS, CPA, impressions by campaign/ad
* Campaign/ad status (active, paused, limited, disapproved)
**Analysis approach**
1. Calculate DoD and WoW % changes.
2. Compare to rolling averages (14d+) to separate normal variance from unusual spikes/drops.
3. Use **statistical significance and impact weighting** (spend + conversions volume) to decide what's meaningful — not fixed thresholds. Look for both:
* **Sudden anomalies** (sharp spikes/drops).
* **Sustained trends** (gradual decline or improvement across 5+ days).
* **Cross-campaign shifts** (one campaign down, another up, account stable).
* **Silent delivery loss** (impressions collapse while budgets unchanged).
4. Diagnose causes by considering:
* **Tracking/attribution issues** (conversion loss with steady spend/clicks).
* **Budget/spend issues** (pauses, caps, reallocations).
* **Approval/delivery issues** (disapprovals, "limited" status, audience exhaustion).
* **Auction/competition dynamics** (CPC/CPM spikes, impression share changes).
* **Creative fatigue** (CTR drops).
* **External context** (holidays, promotions, platform outages).
**Interpretation guidance**
* If conversions drop suddenly → check tracking, site, attribution first.
* If spend = 0 → check budgets, pauses, disapprovals.
* If spend spikes but conversions don't → wasted reach, targeting/settings.
* If CPC/CPM rises but CTR flat → auction pressure.
* If CTR drops but CPC steady → creative fatigue.
* If impressions collapse with steady budgets → delivery issue or disapproval.
* If gradual 5+ day decline → structural issue (auction, creative, competitor surge).
**Output format must be:**
1. **Coverage check** – # campaigns reviewed, % with anomalies.
2. **Table of flagged fluctuations**, columns:
* Campaign/Ad
* Metric (spend, conversions, CPC, CTR, impressions)
* % change (DoD/WoW or sustained trend)
* Impact level (High / Medium / Low)
* Likely cause (tracking, budget, approval, auction, creative, external)
* Recommended action (urgent, optimise, monitor)
3. **Grouped recommendations:** Do Now / Do Next / Monitor.
4. **Plain-language summary** (e.g. "Campaign A conversions -65% with clicks steady → likely tracking issue. Campaign B impressions -45% → possible disapproval. Campaign C CPC +30% with CTR flat → auction pressure.").
Be concise, structured, and action-oriented.
Creator
Stay Updated
Get AI marketing tips and template updates delivered to your inbox.
Creator
Stay Updated
Get AI marketing tips and template updates delivered to your inbox.
