← Guides·Google Analytics 4

Google Analytics 4

GA4 alerts: how to set them up and where they fall short

How to create GA4 alerts with custom insights, which conditions and frequencies to use, and the tracking failures a metric alert never sees.

Rafael CampoamorDigital Marketing Freelancer
6 min read · 1293 words

GA4 alerts are called custom insights. You set a condition on a metric, such as purchases down more than 20% or revenue with an anomaly, GA4 checks it on the schedule you choose, and it tells you by email and on the Insights dashboard when the condition is met. They catch big moves in your reports.

They watch the report, not the tracking. That difference decides what they can catch.

A smoke detector mounted on a white ceiling
An alarm only reacts to what reaches it. Image: Curpharar, Wikimedia Commons, CC BY-SA 4.0

What a GA4 alert is

Google groups alerts under Analytics Intelligence, described as features that use machine learning and conditions you configure (Google’s guide to analytics insights). There are two kinds:

SettingWhat GA4 offers
Evaluation frequencyHourly (web data only), daily, weekly, or monthly
ConditionA metric crossing a value or changing by a percentage, or “has anomaly”
SegmentAll users by default, or include or exclude a dimension value
NotificationInsights dashboard for everyone on the property, email to the addresses you add
Limits50 custom insights per property, kept for one year, Editor or Administrator role to create them

How to set up a GA4 alert

  1. From the GA4 home page, scroll to the Insights section and click View all insights, then Create.
  2. Pick a suggested insight or click Create new to start from scratch.
  3. Choose the evaluation frequency. Pick hourly only for web data that moves within a day, like purchases or sign-ups.
  4. Set the segment. Leave it on all users, or narrow it, for example to one country or one traffic source.
  5. Set the condition: a metric, a condition such as “% decrease more than”, and a value. Or choose has anomaly and let GA4 decide what counts as unusual.
  6. Name the insight, add the email addresses of the people who should hear about it, and save.

You can switch email notifications on or off for each insight later, under Manage on the Insights dashboard.

Which alerts are worth creating

With 50 slots and an inbox to protect, start with the metrics that cost money when they move.

AlertFrequencyCondition
Purchases dropHourly (web) or dailyPurchases, % decrease more than a level you pick from normal daily swings
Revenue anomalyDailyPurchase revenue, has anomaly
Lead or sign-up dropDailyYour key event count, % decrease more than your threshold
Paid traffic dropDailySessions, segment on paid channels, has anomaly
Checkout start collapseDailybegin_checkout event count, % decrease more than your threshold
Traffic spikeDailySessions, % increase more than your threshold, to catch bots and duplicate tags

Set thresholds from your own history, not round numbers. Look at a few weeks of the metric by day of week and choose the drop that would never happen on a normal day. If an alert fires every week and nobody acts on it, raise the threshold or delete it.

Where GA4 alerts fall short

A custom insight evaluates the numbers GA4 reports. Every limitation follows from that.

What a GA4 alert can and can’t see: it watches reported metrics such as purchases and revenue after processing, while breaks upstream, such as a missing parameter, a pixel that stopped firing, a tag before consent, or a duplicate event, may never move the metric it watches.
What a GA4 alert can and can’t see: it watches reported metrics such as purchases and revenue after processing, while breaks upstream, such as a missing parameter, a pixel that stopped firing, a tag before consent, or a duplicate event, may never move the metric it watches.

They see missing data only when it moves the metric. If a release removes the purchase event from one checkout step, purchases fall and an alert on purchases may fire, but it can’t tell you the event is missing or why. If the release removes the transaction_id or the currency parameter, the purchase count stays flat and that alert stays quiet. A revenue alert might move, and deduplication breaks without any metric showing it.

They only see GA4. A Meta Pixel that stopped sending purchases, a Google Ads conversion tag that lost its value, or a TikTok event firing twice won’t move a GA4 metric at all.

They don’t separate demand from tracking. A 30% drop in purchases looks the same whether the campaign stopped selling or the tag stopped firing. The alert tells you a number moved, not why, which tag, or which release.

They wait for the next evaluation. A daily alert looks at a whole day of data, so a break on Monday morning may not show until the next evaluation. Hourly is faster, but only for web data.

You pick what to watch. You choose the metric, the segment and the schedule for each insight. With “has anomaly”, GA4 decides what counts as unusual; with any other condition, you set the threshold and maintain it as traffic changes. Fifty insights cover the headline metrics, not every event and parameter.

Consent and privacy are out of scope. A tag that fires before consent, or an email address leaking into a URL parameter, changes no metric GA4 alerts watch.

For the wider set of monitoring, alerting, and logging practices these alerts sit inside, Google Cloud’s Architecture Framework is a useful frame, and the Apigee documentation shows the same threshold model applied to APIs.

How to cover what GA4 alerts miss

Keep GA4 alerts for the business metrics. Add monitoring at the level of the hit for everything else: every event, every parameter, and every destination, checked against what normally arrives.

That is what Trackingplan does. It learns your baseline from real traffic during the first days, per hour of day and day of week, then compares every hour against it, per event, per property, and per platform. It posts only the deltas that matter, with an explanation and links to the affected hits, in the channel that owns it. Beyond volume, it catches missing properties, dead tags, duplicate fires, broken triggers, and consent breaches, across GA4 and the ad pixels alike.

The two answer different questions. A GA4 alert says purchases fell. Monitoring says which event lost which parameter after which release. For a side-by-side comparison, see Trackingplan vs GA4 alerts. For anomaly detection inside GA4 itself, see GA4 anomaly detection.

If your alerts are part of a wider GA4 setup, the GA4 ecommerce guide covers the purchase events these alerts depend on, and the rest of our GA4 guides go further into advanced data analysis in GA4 and customer journey analysis for ecommerce.

FAQ

Does GA4 have alerts?

Yes. In GA4, alerts are called custom insights. You create a rule on a metric, GA4 evaluates it hourly, daily, weekly, or monthly, and when the condition is met the insight appears on the Insights dashboard and, if you add addresses, goes out by email.

How many custom insights can a GA4 property have?

Up to 50 per property. Triggered insights are kept for one year, and creating or editing them needs the Editor or Administrator role.

Can GA4 alert me when tracking breaks?

Sometimes, indirectly. A custom insight watches a reported metric, so it fires when a tracking break moves that metric far enough. A missing parameter can leave the event count unchanged even when it affects another metric, a pixel on another platform never reaches GA4, and the insight can’t say which tag or release caused the change.

Are hourly GA4 alerts available for apps?

No. Hourly evaluation works only for web data. For app data, Google offers daily, weekly, or monthly evaluation, because delayed app events could trigger false notifications.

Rafael CampoamorDigital Marketing Freelancer

Digital Marketing Freelancer

Trackingplan

See everything. Miss nothing.

Your implementations audited around the clock with real-time, real-user data. Real-time alerts about errors or changes in your data, campaigns, pixels, privacy and consent. Let AI flag issues before they cost you.