How to Set Up Anomaly Detection for Your Website Traffic
By Emily Redmond, Data Analyst at Emilytics · April 2026
TL;DR: Anomaly detection automatically alerts you to unusual traffic spikes or drops. Instead of manually checking GA4 daily, set it up once and get notified when something changes. Game-changer for catching issues fast.
What is Anomaly Detection?
Anomaly detection uses AI to learn your "normal" traffic pattern, then alerts you when something looks unusual.
Examples:
- Your normal daily traffic: 1,000 users. On Thursday, you get 5,000. Alert: "Traffic spike detected."
- Your normal conversion rate: 2%. On Friday, it drops to 0.8%. Alert: "Conversion rate drop detected."
- Your normal bounce rate: 48%. On Monday, it jumps to 78%. Alert: "Bounce rate anomaly."
The system learns from historical data and accounts for seasonality, so it won't alert you every time.
Why You Need Anomaly Detection
Manually checking GA4 daily is:
- Time-consuming
- Error-prone (you might miss subtle drops)
- Reactive (you find out problems after they've been happening for hours)
Anomaly detection is:
- Automatic
- Accurate (based on statistical models)
- Proactive (alerts within hours)
A traffic drop caught in 2 hours is way more recoverable than one you notice 2 days later.
Setting Up Anomaly Alerts in GA4
GA4 has built-in anomaly detection in the Insights feature.
Step 1: Enable Insights
- In GA4, go Admin (left sidebar, bottom)
- Under Account, select Account Settings
- Look for Insights and make sure it's enabled
- Note: This is a GA4 standard feature (free, no extra charge)
Step 2: Create an Anomaly Alert
- Go Reports → any report (e.g., Acquisition → Overview)
- Look for Insights or an alert icon (looks like a light bulb or notification bell)
- Click to create a new insight/alert
- Choose:
- Metric: Sessions, Users, Conversions, Bounce Rate, etc.
- Condition: Increases by more than X%, Decreases by more than X%
- Dimensions: Optional (e.g., only alert if drop is in organic search)
Example Alert: Alert me if daily sessions decrease by more than 20% compared to recent pattern.
Step 3: Set Email Notifications
- In the alert, set Email notifications to ON
- Add your email address
- Choose frequency:
- Daily: Get an email if anomaly detected (summarized)
- Real-time: Get alerted immediately
Real-time is better for urgent metrics (conversions, major traffic sources).
Using GA4 Insights
GA4's Insights feature automatically detects anomalies for you.
- Go Reports → Acquisition → Overview
- Look for Insights section (usually at bottom or top)
- GA4 will show automatic insights like:
- "Sessions increased 35% above usual on Mar 18"
- "Bounce rate unusually high on Mar 17 (77% vs usual 45%)"
These are automatic anomalies GA4 detected. Click to investigate.
Advanced: Using Emilytics for Anomaly Detection
Emilytics is an AI agent that watches your GA4, Google Search Console, and Bing data 24/7 and alerts you to anomalies.
Benefits over GA4's built-in:
- Monitors all sources (GA4 + GSC + Bing) in one place
- AI analyzes root cause ("Your traffic dropped because you lost ranking for 'X keyword'")
- Faster alerts (real-time vs. daily)
- Better anomaly detection (learns your patterns over time)
How to Use Emilytics
-
Connect your GA4 and Google Search Console accounts
-
Emilytics automatically monitors traffic 24/7
-
When an anomaly is detected, you get an alert with context:
- "Traffic down 22% from 'organic search'"
- "Top ranking for 'best practices' dropped from #1 to #3"
- "Direct traffic spike from (unknown country)"
-
Click the alert to see root cause analysis and recommended fixes
Common Anomalies to Alert On
Anomaly 1: Traffic Drop (Critical)
Alert if daily traffic drops more than 25%.
Why? Catches ranking loss, site issues, campaign stops, bot filtering changes.
Example alert: "Sessions decreased 28% on March 15 (from 1,200 to 860)"
Anomaly 2: Conversion Rate Drop (Critical)
Alert if conversion rate drops more than 30% (relative).
Why? Catches checkout broken, form issues, landing page problems.
Example alert: "Conversion rate decreased 35% on March 15 (from 2.1% to 1.4%)"
Anomaly 3: Traffic Source Anomaly (High)
Alert if a major source's traffic drops more than 20%.
Why? Catches organic search drop, ad campaign stop, referral loss.
Example alert: "Organic search traffic decreased 32% on March 15"
Anomaly 4: Bounce Rate Spike (Medium)
Alert if bounce rate increases more than 15 percentage points.
Why? Catches mobile UX break, landing page broken, messaging mismatch.
Example alert: "Bounce rate increased to 62% on March 15 (from usual 47%)"
Anomaly 5: New Traffic Source (Low)
Alert if traffic from an unknown or unusual source appears.
Why? Catches bot traffic, click fraud, new referral source.
Example alert: "Traffic from unfamiliar source 'x.x.x.x' on March 15 (500 sessions)"
Investigating an Anomaly When Alerted
When you get an anomaly alert, follow this process:
Step 1: Confirm It's Real
Open GA4. Navigate to the metric that triggered the alert. Is it really abnormal?
Sometimes alerts fire on false positives (normal variation).
Step 2: Check the Obvious
- Did I launch a campaign on that date?
- Did I pause a campaign?
- Did I change GA4 settings?
- Did I launch new ads?
- Did I publish content?
Step 3: Dig Into the Metric
For a traffic drop:
- Which source dropped? (Organic? Paid? Social?)
- Which pages lost traffic?
- Did bounce rate increase? (Suggests traffic quality issue)
For a conversion drop:
- Did all conversion types drop, or one specific type?
- Did bounce rate increase? (Suggests traffic changed)
- Did conversion rate drop on all pages, or specific pages?
Step 4: Diagnose
Based on what you found, figure out the root cause:
If organic traffic dropped:
- Check Google Search Console: Did rankings drop?
- Check top landing pages: Lost traffic on all or specific pages?
- Is this seasonal? (Compare to same date last year)
If paid traffic dropped:
- Check your ad account: Did you pause campaigns?
- Check daily budget: Did it run out?
If conversion rate dropped:
- Check landing pages: Did conversion rate drop on all pages?
- Check traffic sources: Did traffic source change (e.g., lower-quality source)?
- Check forms: Are forms broken? (Check console errors)
Step 5: Act
- If campaign paused accidentally: Restart it
- If site broken: Fix it immediately
- If seasonal: Don't panic, monitor
- If ranking dropped: Start SEO work to recover
- If tracking broken: Fix tag implementation
Setting Alert Thresholds
Too strict = too many false alerts (you ignore them).
Too loose = you miss real problems.
Recommended thresholds:
| Metric | Increase Alert | Decrease Alert |
|---|---|---|
| Sessions | > 50% | > 25% |
| Conversions | > 75% | > 30% |
| Bounce Rate | > 15 pp | No alert needed |
| Conversion Rate | No alert needed | > 30% |
| Traffic Source (organic) | No alert needed | > 20% |
Start with these. Adjust after a few weeks based on false positive rate.
Frequently Asked Questions
Q: How often should I check my alerts? A: Set email frequency to daily or real-time (if critical metrics). Check inbox every morning. Act within 2 hours if it's a real issue.
Q: Can I get alerts on specific pages? A: GA4's built-in alerts are limited. Use Emilytics or set up custom alerts in Google Sheets/Data Studio if you need page-level anomalies.
Q: What if I'm getting too many alerts? A: Adjust thresholds. Make them less strict (20% drop instead of 10%). Or filter to only critical metrics (conversions, major sources).
Q: Should I alert on spikes or only drops? A: Alert on both. Spikes might be bot traffic (bad). Drops are obvious problems (bad). Both matter.
Q: Can anomaly detection catch bot traffic? A: Partially. If bot traffic is abnormal (not your usual pattern), anomaly detection will catch it. If bots visit consistently, it won't. Use GA4's bot filter to clean data first.
The Bottom Line
Set up anomaly detection once. It pays for itself the first time it catches an issue before your boss notices.
Don't rely on manual checks. Automate it. Stay alert. Respond fast.
Emily Redmond is a data analyst at Emilytics — AI analytics agent watching your data around the clock. 8 years experience. Say hi →