Measuring a B2B SaaS trial in GA4: the funnel your reports are missing
A free trial is not a checkout, and the standard GA4 reports treat it as though it were. Someone signs up, the event fires, the session ends, and acquisition reporting moves on to the next visitor. The conversion that actually pays for the campaign happens two or three weeks later, often on a different device, often through a colleague who never saw the ad.
The reports that ship with Analytics are built for sessions and events. A trial is a sequence that spans sessions, devices and sometimes several seats inside one account. Funnel exploration counts users through an ordered set of steps, so its numbers will not match the Acquisition page.
Why the trial disappears from standard reports
An Events report tells you how many times something happened. An Acquisition report tells you which session brought the user in. Neither tells you how many of the people who signed up reached the step where the product starts to feel useful.
Funnel exploration does count that, but it counts it in a particular way. Each user can enter a funnel once, and only the first sequence of steps is reported. If the same person completes the funnel twice in the date range, the second run is ignored. That rule matters for trials, because a person who abandons a trial and starts another one a month later will appear once, at the first attempt, not twice.
The same rule explains the gap between users and accounts. A funnel counts people. A finance team counts contracts. Between those two numbers sit seats, invitations and procurement, and no part of GA4 knows about any of them.
Send the events you intend to count
Google keeps a list of recommended events, sent with prescribed parameters so that predefined dimensions and reports populate without extra work. A few of them cover most of a trial: sign_up (a user signs up for an account), login (a user logs in), purchase (a user completes a purchase) and tutorial_begin plus tutorial_complete for onboarding steps you want to treat as standard.
The method parameter is worth the five minutes it takes. Send it with sign_up and login and the Method dimension separates email registration from Google sign-in, which is the first split most SaaS teams want and rarely have.
The activation step has no recommended event, because activation means something different in every product. Define it yourself: a project created, a first report run, an integration connected, a teammate invited. Send it as a clearly named event rather than inferring it from page views. A funnel step can be built on an event or on a dimension value, for example the dimension User acquired campaign equals a specific campaign, or the purchase event with a value above a threshold. It cannot be built on a metric, so a condition like revenue greater than 500 is not available as a step.
Open or closed, and why it changes the answer
Every funnel is either open or closed. In an open funnel users can enter at any step. In a closed funnel they must enter at the first step. In both cases the steps have to occur in the specified order, and a user who skips a step falls out and is not counted in anything after it.
Google's own worked example is the clearest way to see the difference. Take three steps A, B and C, and four users. One does A, B, C. One does B, C. One does A then C. One does only C. An open funnel counts all four, but the third user appears only at step A, because they skipped B and fell out. A closed funnel counts only the two users who began at A.
For a trial, the headline rate needs a closed funnel. Everyone starts at sign_up, so the closed shape answers the question the business asks: of the people who started a trial, how many reached activation and how many paid.
An open funnel earns its place as a diagnostic. If invited teammates land straight on an invitation acceptance page and never see the signup form, a closed funnel will not show their journey at all. Build the open version as a second exploration when the team keeps asking where invited users get stuck.
Ten steps is the ceiling. That limit is useful: a trial that needs more than ten steps to describe is usually a trial with an unclear definition of activation.
Put a clock on the steps
Each step after the first can be set to follow the previous one directly or indirectly. A step that follows directly has to happen immediately after the previous event. A step that follows indirectly allows other actions in between, which is what almost every product step needs.
Indirect steps accept a Within time frame. Set it. A trial funnel read without one will happily count someone who signed up in January, came back in March and activated on a whim, and that completion will make activation look healthier than it is. For a 14 day trial, a Within of 14 days on the activation step and a wider window on the purchase step is a defensible starting point.
Signup to activation inside one trial is a product question. Signup to paid subscription, including a sales cycle, is a revenue question. Those two funnels should be separate explorations with separate windows.
Reading the step table without fooling yourself
Four details in the output catch people out.
Breakdown dimensions use the first value. If you break the funnel down by device category, a user who enters on mobile and completes every later step on desktop appears in the mobile row throughout, because users are attributed to the first instance of the breakdown value that applies to them. The same rule applies to trended funnels, where users are attributed to the first date they were active in each step.
Elapsed time is an average over completers. Turn on Show elapsed time and GA4 reports, for each step, the total time between the previous step and this one divided by the number of users who reached this step. Only the first completion of a repeated event counts. Averages hide the long tail, so a step that averages four days can still have half its users finishing in an hour and half finishing on day thirteen.
Next action is not always a drop-off. The Top 5 Next Actions card shows what users did after each step. A large (no next action) entry using a page-related dimension often means the page title did not change between steps, not that users stopped. Read the card with the dimension in mind before treating it as an exit.
Filters and segments are different tools. Filters restrict the data to conditions that all have to be met. Segments let you compare up to four groups side by side, and you can create one by right-clicking a row in the data table.
The counting method decides the conversion number
Key events have a counting method, and it changes every conversion figure that follows. Once per event counts the event every time it is triggered. Once per session counts it once inside a session. Google's example is direct: trigger the same key event five times in one session and Once per event reports five, Once per session reports one.
The default depends on how the key event was created. Once per session is the default for key events created from Universal Analytics goals, in an automatically created property or through the goals migration tool in the Setup Assistant after April 2023. Once per event is the default for everything else. The setting can be changed at any time under Events in the Data display menu, and it needs Marketer or above at property level.
For a trial, set it deliberately, event by event. A purchase should count once. An activation step that a user legitimately repeats may be better counted every time, provided the funnel step itself does not then inflate the number of users who reached it. Mixing the two without deciding is how a team ends up with a conversion rate nobody can explain.
Segments that answer a sales question
Once the funnel works, the interesting comparisons are commercial. Compare self-serve trials against sales-assisted trials. Compare trial starts from paid search against trial starts from a partner referral. Compare plan tier, region, or whether the account was created during a campaign period.
Four segments at a time is enough. Pick the two or three splits that would change what the business does next rather than filling every slot.
Set the date range before you read the numbers
A trial funnel read over a period shorter than the trial shows a cliff that does not exist. If the trial lasts 14 days and you look at the last seven days, every recent signup is still mid-trial, so the later steps look empty and the conversion rate collapses. The same mistake appears when purchase is counted inside the trial window but the deal closes after it.
The window needs to cover the full trial plus the lag to purchase, and the report needs a note saying so. Save the exploration as a funnel report when it is right, since a saved report keeps the steps, the window and the conditions together and stops next month's version from being rebuilt from memory.
What the funnel cannot tell you
GA4 counts users, not accounts. A five-seat team trial can produce five users, one buyer and a single contract, and the funnel has no concept of the relationship between them. If account-level reporting matters, the account identifier belongs in your CRM or warehouse, joined to GA4 on a user identifier, not inferred from a funnel.
The purchase step only exists if the purchase happens where the tag can see it. Deals closed by an account executive in a contract tool will not fire purchase, so the final step will read as zero for exactly the customers who pay the most. Import those outcomes from the CRM, or report the funnel to activation and treat revenue as a separate exercise.
Users who decline analytics storage do not appear in the funnel at all. That is not a bug, and it is not something a funnel can model. It does mean the drop-off you see at the first step includes people whose data was never collected, so the first step is the least trustworthy one in the chart.
Build order
- Confirm the four events exist and carry the parameters you need:
sign_up,login, your activation event, andpurchasewhere it applies. - Define activation in one sentence, with the event name that proves it.
- Build a closed funnel from signup to activation, with Within set to the trial length.
- Build a second, open funnel that includes the invitation path for teammates.
- Set the counting method on each key event and write down the choice.
- Fix the date range to cover trial plus purchase lag, then save the report and add a one-line note about the window.
The funnel will not match the Acquisition report, and it should not. One counts sessions and events, the other counts people moving through a sequence. When someone asks why the numbers differ, the answer is in the counting rules, and it is worth writing those rules down beside the chart.
Running a SaaS trial and unsure which step is losing the deal? Book a free tracking audit and we will map the events, check the counting method and build the funnel with your own data.