How is setting up GA4 for SaaS different from the default setup?

GA4's default events and recommended-event library are built for ecommerce: view_item, add_to_cart, begin_checkout, purchase. A SaaS has none of those. Instead you have to define your own event model around the subscription journey — sign_up, an activation event for your product's 'aha' moment, trial_started, key_feature_used, and subscribe (with value and plan parameters). Then you mark the ones that represent success (typically sign_up, activated, subscribe) as Key events. If you just install GA4 and follow a generic tutorial, you'll end up tracking a checkout that never happens and miss the events that actually describe whether users reach value.

What is an activation event in GA4 and why does it matter for SaaS?

An activation event marks the moment a new user first experiences the core value of your product — connecting a data source, sending a first message, creating a first project, whatever proves the product works for them. GA4 doesn't define it for you because it's specific to your product. It matters because it's the single best predictor of whether a signup becomes a paying, retained customer. Without it you can measure traffic and signups, but you can't answer the question that decides which channels are worth scaling: do the users from this source actually reach value? If you instrument only one custom event in GA4, make it your activation event.

Can GA4 track traffic from AI assistants like ChatGPT and Claude in 2026?

Yes. Since May 2026, GA4 resolves referrals from AI assistants — ChatGPT, Gemini, Claude, Perplexity and others — into a dedicated 'AI assistant' channel group, instead of dumping them into Direct or Referral where they were effectively invisible. View it under Reports → Acquisition → Traffic acquisition with the primary dimension set to Session default channel group, and look for the 'AI assistant' row. For B2B SaaS in 2026 this is the first hard signal of how much buyer discovery is happening inside AI answers rather than classic search.

What is a Key event in GA4?

A Key event is GA4's term for what used to be called a 'conversion' (Google renamed it in 2024). You mark an event as a Key event under Admin → Events to tell GA4 it represents a meaningful business outcome. For SaaS, the usual key events are sign_up, activated, and subscribe. The common mistake is marking too many events — a page_view or pricing-page visit is interest, not value. If everything is a key event, none of them are useful for measuring real progress.

What are the most common GA4 setup mistakes for SaaS teams?

Five recurring ones: (1) leaving data retention at the 2-month default instead of pushing it to 14 months, so you lose year-over-year analysis; (2) marking page_view or pricing visits as conversions instead of real progress events; (3) never defining an activation event, so you can't tell which channels produce users who reach value; (4) defining an internal-traffic filter but forgetting to switch it from Testing to Active, which inflates engagement; and (5) treating the GA4 UI as ground truth once you scale, when sampling and the (other) row distort it — the BigQuery export is the version that doesn't lie.

How to Set Up GA4 for SaaS in 2026: Events, Key Events, BigQuery & AI Assistant Traffic

Q: Should I enable GA4 BigQuery export for a SaaS?

Yes, and you should do it on day one. The free GA4 → BigQuery export sends raw, unsampled event rows to a warehouse you can query with SQL — it used to cost six figures a year on Universal Analytics 360 and is now free for every GA4 property. Critically, there is no backfill: the export only captures data from the moment you enable it forward, so turning it on early is the only way to have history later. As you scale, the GA4 reporting UI and Data API start distorting data through sampling, data thresholds, and the '(other)' row, while BigQuery stays exact. Enable it under Admin → Product links → BigQuery links.

Google Analytics 4 was designed for retail. The default events assume a cart, a checkout, and a purchase. None of those exist in a SaaS funnel — yet most founders install the GA4 snippet, see numbers start flowing, and assume they're measuring the right things. They're not. They're measuring page views and a checkout that never happens.

This is the setup we'd give a B2B SaaS founder starting GA4 from scratch in 2026: what to configure on day one, the event model that actually maps to a subscription business, and the two GA4 changes new this year that matter most for SaaS.

GA4 out of the box answers “did they buy?” A SaaS needs it to answer “did they sign up, activate, subscribe, and stick?” — and that's a model you have to build.

Why the default GA4 setup misleads SaaS teams

GA4's enhanced measurement fires page_view, scroll, session_start, and a handful of others automatically. Its recommended-event library — the stuff every tutorial walks you through — is built around view_item, add_to_cart, begin_checkout, and purchase. That's an ecommerce skeleton.

A SaaS doesn't convert in a single transaction. It converts across a sequence: someone signs up, reaches the “aha” moment that proves the product works, starts a trial, uses the core feature, and eventually subscribes — then hopefully invites a teammate and upgrades. If GA4 only knows about page views and a phantom checkout, it can tell you how many people arrived, but nothing about whether they ever reached value.

Fig. 1 — The events GA4 ships with describe a store. The events that describe a subscription business — sign_up, activated, subscribe, retained — are ones you have to define yourself.

The fix isn't complicated, but it has to be deliberate. You define the events that matter for your funnel, send them with meaningful parameters, and tell GA4 which ones are the conversions you care about. Everything downstream — reports, audiences, attribution — depends on getting that model right first.

The six things to set up on day one

Before you trust a single number, fix these. Three of them (data retention, BigQuery export, and your key events) are effectively irreversible if you skip them — you cannot backfill data you never collected.

Fig. 2 — The day-one checklist. The exact menu path is listed for each so you can do all six in one sitting.

1. Push data retention to 14 months

GA4 defaults event-level data retention to 2 months. For any analysis beyond “last quarter,” that's useless — you lose year-over-year comparisons before you ever run one. Set it to the maximum 14 months immediately under Admin → Data settings → Data retention. This only affects the Explore/exploration data; standard aggregated reports keep history longer, but you want the granular data too.

2. Define your SaaS key events

These are the events worth instrumenting for almost any B2B SaaS, with the parameters that make them useful later:

sign_up — with a method parameter (google, email, sso) so you can see which signup path converts best.
activated — the single most important event, and the one no tutorial defines for you because it's specific to your product. It's the moment a user experiences core value (connected a data source, sent a first message, created a first project). If you instrument one custom event, make it this one.
trial_started and key_feature_used — the in-between signals that tell you whether activation is sticking.
subscribe — with value and plan parameters, so GA4 can attribute real revenue to a channel.
invited_teammate, upgraded — expansion signals that matter for product-led growth.

Then mark the ones that represent success as Key events (GA4's 2024 rename of “conversions”) under Admin → Events. For most SaaS teams that's sign_up, activated, and subscribe. Resist the urge to mark everything — if every event is a conversion, none of them are.

3. Turn on BigQuery export — today, not later

The free GA4 → BigQuery export sends raw, unsampled event rows to a warehouse you can query with SQL. It used to cost six figures a year on Universal Analytics 360; on GA4 it's free for every property. The catch: there is no backfill. The export only captures data from the moment you enable it forward. Switch it on under Admin → Product links → BigQuery links on the same day you create the property, even if you won't write a query for months.

Once you're actually analyzing data, the GA4 reporting UI and Data API will start lying to you in subtle ways — sampling, data thresholds, and the dreaded “(other)” row. We wrote up exactly how and why in Stop Using the GA4 API — Use BigQuery Export Instead. The short version: the warehouse is the source of truth; the UI is a preview.

4. Filter internal and developer traffic

Your own team logging in, your QA clicks, and your localhost testing all inflate engagement and pollute activation rates. Define an internal-traffic filter (by IP or a traffic_type parameter) under Admin → Data streams → Configure tag settings → Define internal traffic, then activate it under Data settings → Data filters — it ships in “Testing” mode and does nothing until you switch it to Active. This is the step people configure and forget to turn on.

5 & 6. Link Search Console and Google Ads

GA4 cannot show you the search queries that brought organic visitors — that data only lives in Google Search Console. Link them under Admin → Product links → Search Console links so query data appears inside GA4 reports. If the two never seem to agree afterward, that's expected, not a bug — see GA4 vs. GSC: Two Tools, One Funnel, Endless Confusion for which one to trust for which question. Likewise, link Google Ads to close the loop between spend and signups.

What's new in GA4 for 2026 (and why it matters for SaaS)

GA4 shipped a run of changes over the last six months that matter specifically for SaaS teams trying to understand where buyers actually come from. Two are worth your attention during setup — both are on automatically, so this is about knowing where to look, not flipping a switch.

AI assistant traffic is now its own channel

Since May 2026, GA4 resolves referrals from AI assistants — ChatGPT, Gemini, Claude, Perplexity, and others — into a dedicated “AI assistant” channel group instead of dumping them into Direct or Referral where they were effectively invisible. For a B2B SaaS in 2026, a growing share of buyers ask an AI engine before they ever open Google, so this row is the first hard signal of how much of your discovery is happening inside AI answers.

Fig. 3 — Find it under Reports → Acquisition → Traffic acquisition, with the dimension set to Session default channel group. The 'AI assistant' row is the traffic that used to hide inside Direct.

Check it under Reports → Acquisition → Traffic acquisition and set the primary dimension to Session default channel group. If the AI assistant row is non-trivial and growing, that's your cue to take generative engine optimization seriously.

Analytics Advisor and Generated Insights

Google added an AI assistant inside GA4 (“Analytics Advisor,” rolled out over late 2025 into 2026) that answers questions about your data in plain language, plus Generated Insights on the Home page that auto-summarize the top changes since your last login. They're genuinely useful for a quick pulse-check, with one honest caveat: they read whatever you configured. If your event model is the ecommerce default, the AI will confidently summarize the wrong funnel. Get steps 1–2 right first, or the insights inherit the same blind spots.

Common SaaS setup mistakes

Marking page_view as a conversion. A pricing-page visit is interest, not value. Conversions should be the events that represent real progress — sign_up, activated, subscribe.
No activation event. Without it, you can measure traffic and signups but never answer the question that decides whether a channel is worth scaling: do the users from this source actually reach value?
Sending revenue without a currency or value. A subscribeevent with no value parameter can't attribute revenue to a channel. GA4 will happily report “0” conversions value and you'll trust it.
Forgetting to activate the internal filter. Defining it isn't enough — it ships in testing mode. Half of all “why is my engagement so high” questions trace back to this.
Treating the GA4 UI as ground truth. Once you scale, sampling and the (other) row distort the reports. The BigQuery export is the version that doesn't lie.

Where Alice fits

A correct GA4 setup gets you the data. The harder part is the standing question every week: which signups activated, which channel actually drives subscriptions, and whether the AI assistant row is quietly becoming your biggest source of qualified buyers. That's reconciliation work across GA4, Search Console, Google Ads, and Stripe — and it's the work most founders never have time for.

Alice, the AI growth analyst by TranX, connects your GA4 (including the BigQuery export), joins it to your other sources, and lets you ask the questions in plain English — “which channel has the highest activation-to-paid rate this quarter?” — with the real, unsampled numbers and the query shown so you can verify it. Set GA4 up properly first; let Alice do the watching.

How to Set Up Google Analytics 4 for SaaS (2026)