The best way to get GA4 alerts in Slack without building a Zap

The best GA4 alert in Slack is one that almost never fires. When it does, it should say what is unusual compared with a normal day of the same kind, and land in the channel of the person who can act on it. GA4 has no Slack integration, so you have four ways to get there: a custom insight email sent to a channel address, a script on the GA4 Data API, an automation tool like Zapier, Make or n8n, or a dedicated tool. Only the last two can judge what “normal” means for you, and only if you build it or buy it.

This article is about alerts, the messages that interrupt you when something breaks. For the Monday summary, see how to send a GA4 weekly report to Slack. For the whole picture, the GA4 in Slack guide covers both.

Why most GA4 alerts in Slack get muted

Alert channels die the same way. People read every message the first week, someone mutes the channel after the third false alarm, and the next real drop sits there unread.

The false alarms come from four causes:

• The wrong baseline. Weekend traffic is lower than weekday traffic for most B2B sites. A rule that compares yesterday with the day before fires every Saturday, and again every Monday when traffic bounces back.
• No minimum volume. A property with 8 conversions a day will swing 50% on pure chance. A percentage rule on a small number is a random number generator.
• No cause. “Sessions down 32%” with nothing else sends someone into GA4 to find out whether it is a tracking bug, a campaign that ended, or a public holiday. If the answer is boring most of the time, people stop checking.
• The wrong channel. If you run ten clients, one shared #alerts channel means everyone sees everything and nobody owns anything.

What makes a useful traffic or conversion alert

Fix those four causes and you have the spec.

1. Compare with the same weekday. A Tuesday is measured against recent Tuesdays, not against Monday or a flat 30-day average. Weekly rhythm stops looking like an anomaly.
2. Require both a statistical surprise and a meaningful size. A move can be statistically unusual and still too small to matter, or large in percentage terms and still normal for a noisy metric. Ask for both, and skip metrics whose typical daily value is too low to measure.
3. Say what is unusual, in numbers a person can check. “1,240 sessions yesterday against about 1,900 on a typical Tuesday” beats “anomaly detected”. Then make the next question cheap to ask: which channel, which page, since when.
4. Route it to the people who own the account. One channel per client or per site. The account manager sees their client’s alert, and the client can see it too if the channel is shared with them.

A drop on a national public holiday is real but expected, and should not wake anyone up. If you want a checklist for working out the cause once an alert fires, we wrote one: why did my traffic drop in GA4.

Option 1: GA4 custom insights, emailed to a Slack channel

GA4 can already watch your metrics. A custom insight is a rule evaluated hourly, daily, weekly or monthly (hourly for web data only), with either a threshold condition such as ”% decrease more than 20” or a “Has anomaly” condition where GA4 decides what is unusual. You can create up to 50 per property, you need the Editor or Administrator role, and you can enter email addresses to notify.

To get those emails into Slack, use the channel’s own email address. Open the channel, click its name, and on the About tab choose Send emails to this channel. This is available on paid Slack plans only, and workspace admins can turn it off.

Why it works: free, no code, and set up in 15 minutes. The “Has anomaly” condition uses a model trained on 90 days of history for daily anomalies (Google’s documentation), which beats a fixed threshold.

Why it breaks: the alert lands as a forwarded email, with no cause and no way to ask a follow-up in the thread. You maintain the rules property by property. Google’s help pages don’t say how the anomaly model handles day-of-week patterns, so you only find out how noisy it is by running it. We cover the setup and its limits in detail in GA4 custom insights.

Option 2: Apps Script on the GA4 Data API

Write a Google Apps Script that runs every morning on a time-driven trigger. It calls the GA4 Data API for the last few months of daily data, compares yesterday with the same weekday, and posts to a Slack incoming webhook when something crosses your rule.

Why it works: you control the baseline, the minimum volume and the wording. It costs nothing to run.

Why it breaks: you own the statistics and the maintenance. An incoming webhook posts to the single channel chosen when it was installed, so ten clients means ten webhooks to manage. When an OAuth token expires, the script fails without a sound, and silence looks exactly like a quiet day.

Option 3: Zapier, Make or n8n

This is where most “GA4 to Slack” tutorials send you, so check the triggers before you start. As of October 2026:

• Zapier’s Google Analytics 4 app has one trigger, New Conversion Event (“Triggers when a new conversion event is created”), plus actions such as Run Report for a Property. Nothing fires when traffic or conversions move.
• Make’s Google Analytics 4 app lists action modules (Generate a Report, Generate a Pivot Report, Create an Event, Make an API Call) and no trigger.
• n8n’s Google Analytics node offers Report: Get and User Activity: Search, and no trigger.

So every automation-tool “alert” works the same way: a schedule (for example Schedule by Zapier’s Every Day trigger) runs a report, a code or filter step decides whether the numbers are bad enough, and a Slack step posts the message.

Why it works: visual, quick to start, and easy to chain other steps, such as creating a task when an alert fires.

Why it breaks: a filter step is a threshold, and a threshold is the source of the noise described above. A same-weekday baseline with a minimum volume means writing code inside the tool, at which point you have Option 2 with a task bill. Each client is another copy of the workflow to keep in sync.

Option 4: a dedicated tool

The last option is a product that does the detection for you. This is what we built Dialytics for, so read this section knowing it is our product.

Dialytics checks each connected GA4 property once a day. It compares yesterday’s sessions, users, new users and key events with the same weekday over roughly the last 12 weeks, using a median rather than an average so one past campaign spike does not distort the baseline. It speaks only when a move is both statistically unusual and at least 20% away from normal, skips metrics with too little volume to judge, holds back drop alerts on national public holidays, and sends at most one alert a day. On a normal day, the channel stays silent.

When an alert lands, it gives the number and what is typical for that weekday. To find the cause, anyone in the channel mentions the bot in the thread and asks: “which channel drove this?”, “is it one landing page?”. The answer draws on the property’s recent history.

The four options compared

Option	Setup	Weekday baseline	Explains the cause	Per-client channels	Cost
Custom insight to channel email	15 minutes	Not documented	No	One email address per channel	Free, paid Slack plan
Apps Script + Data API	Half a day or more	If you code it	No	One webhook per channel	Free, your time
Zapier, Make or n8n	1 to 2 hours per workflow	If you code it	No	One workflow per client	Tool plan
Dedicated tool (Dialytics)	Minutes	Yes	Through thread Q&A	Yes	Subscription

Which one should you pick?

• One site, a paid Slack plan, low stakes: start with a “Has anomaly” custom insight sent to a channel address. Watch it for a month. If it stays quiet on normal weeks, you are done.
• You have a developer and want full control: Apps Script, with a same-weekday baseline and a minimum volume from day one.
• Your team already runs everything through Zapier, Make or n8n: fine, but put the comparison logic in a code step rather than a simple filter, or the channel will be muted within weeks.
• You run several clients and want alerts you can question in the thread: a dedicated tool saves you from maintaining one workflow per client.

Whatever you pick, judge it by one test: after a normal week, the alert channel should be empty.