How to Find Where Users Drop Off in SaaS Onboarding

Every SaaS onboarding flow leaks. Some new users never verify their email, some never finish setup, and some complete everything and then never return. Leaks are normal. What matters is knowing where the biggest one is, so you spend your limited product time on the fix that recovers the most users. Onboarding drop off analysis for a SaaS product sounds like a job for a data team, but a small team with basic product analytics and a spreadsheet can do it well.

The method below has nine steps. The first five find the leak with numbers, the next three explain it with research and turn it into a fix, and the last checks whether the fix worked.

Step 1: Map the onboarding steps

Start by writing down every step between signup and your activation event, the action that best predicts a user will stick around. A typical B2B flow runs from account creation to email verification, then the first session, workspace or profile setup, connecting an integration or importing data, creating a first core object such as a project or report, inviting a teammate, returning for a second session, and finally reaching activation.

Be specific about your own product. Vague steps such as "explored the product" cannot be measured. If there are several paths through onboarding, map each one. A product that lets users either import their own data or start from a template has two different funnels, and the leak may be in only one of them. Writing the map down also tends to surface steps the team forgot existed, such as a mandatory profile screen that someone added two years ago.

It is worth walking through the flow yourself with a fresh account while you map it. You will notice delays, confusing copy and unexpected screens that do not show up in any spec.

Step 2: Make sure each step is tracked

For each step on the map, confirm there is a reliable event in your product analytics or database. This is less glamorous than analysis and much more important, because tracking gaps look exactly like user drop-off.

A few gaps are especially common. Email verification is often recorded by the authentication provider but never sent to your analytics tool, so it looks like nobody verifies. Integrations frequently track the attempt to connect but not the successful connection, which hides failures. Actions that happen on the server, such as a data import completing in the background, are often missing from front-end analytics entirely.

Fix these gaps before drawing conclusions. If you redesign a step because the funnel showed a steep drop there, and the drop was really a missing event, you have spent weeks solving a problem that did not exist. When in doubt, cross-check the analytics count against your database for a sample of accounts.

Step 3: Build a funnel

Take a recent signup cohort and count how many accounts reached each step. Then calculate the conversion rate between consecutive steps, not just from signup to each step. Step-to-step conversion is what tells you where people leave. Most analytics tools have a funnel report that does this, but a spreadsheet works fine for a few hundred accounts.

A worked funnel (illustrative)

Imagine a cohort of 500 signups. Of those, 430 verify their email, which is 86 percent of signups. Then 400 start a first session, or 93 percent of those who verified. Only 180 connect an integration, which is 45 percent of those who started a session. Of those, 150 create a first report, or 83 percent, and 120 reach activation, 80 percent of those who created a report.

Read step by step, the leak is obvious. Every transition converts at above 80 percent except one. More than half of users who start a session never connect their data, and everything downstream is limited by that. Improving report creation from 83 to 90 percent would help a little. Getting more people through the integration step would help far more, because it is where most of the potential users are lost.

Step 4: Use cohorts and time windows

Measure each step within a defined window, such as seven days from signup, and compare cohorts by week or month. Without a time window, older accounts that eventually completed a step months later inflate the numbers and make recent performance look better than it is. Without cohorts, you cannot tell whether a change you made last month helped.

Choose a window that matches how your product is used. If most activated users get there within three days, a seven-day window is generous enough. If setup typically involves waiting for an IT approval, a longer window may be fairer. Keep the window consistent once chosen, so comparisons between cohorts stay meaningful.

Step 5: Segment the funnel

An average funnel can hide very different stories. Break the funnel down by acquisition channel, company size, the role or use case chosen at signup, device, and plan or trial type. You are looking for segments where one step behaves very differently from the rest.

Segments often change the diagnosis entirely. If signups from paid ads drop off sharply at a step where referral signups sail through, the problem may be traffic quality or the expectations set by the ad, not the onboarding step itself. If mobile users rarely connect integrations, perhaps the integration flow simply does not work well on small screens. If one use case activates far better than others, that might shape both onboarding and marketing.

Be careful with small segments. A segment with twenty signups can swing wildly from month to month. Treat differences in small segments as leads to investigate, not conclusions.

Step 6: Watch what happens at the leak

Numbers tell you where users leave. Qualitative research tells you why. Focus this research on the biggest leak you found, rather than spreading it across the whole flow.

Session recordings

Watch recordings of users arriving at the problem step. Look for repeated clicks on the same element, which often means something looks clickable but is not, or a button that seems to do nothing. Watch for long pauses, which suggest confusion or a decision the user is not ready to make. Notice when users open help documentation and then leave, and when error messages appear. Ten or fifteen recordings usually reveal a pattern.

User tests

Ask a few people who match your audience to complete onboarding while thinking aloud. Recordings show what happened; tests let you hear what people were thinking at the moment they hesitated. Even three or four sessions can be revealing, especially if the participants have never seen your product.

Support tickets

Search your support inbox for tickets mentioning the step. Recurring questions such as "where do I find my API key" or "why does the connection keep failing" point directly to friction.

A short in-app survey

Trigger a single question for users who start the step and abandon it, for example:

What stopped you from connecting your data today?

Keep it to one question with an optional text box. Responses will be few, but the ones you get are from exactly the people you are trying to understand.

Step 7: Interview users who left

Email a handful of users who dropped off at the leak and ask for fifteen minutes. Some will agree, especially if you are clear that you are trying to improve the product, not sell them anything. Their answers often fall into a few patterns. Some needed admin access they did not have. Some were unsure what would happen to their data if they connected it. Some simply got busy and forgot. Others discovered the product was not what they expected.

Each of those answers points to a different fix, which is why interviews matter. Without them, a team tends to assume the cause is complexity and simplify the step, when the real problem was permissions or trust.

Step 8: Prioritize fixes

Match each cause to a fix. When the problem is permissions, let users invite an admin or colleague to complete the step on their behalf, and let them keep exploring in the meantime. When the problem is trust, explain on the step itself exactly what data you access, what you do with it and how it is secured, in plain language. When the step is too complex, simplify it, or offer sample data so users can see the product working before they commit their own. When users simply forget, send a reminder triggered by their behavior, such as an email a day after they started but did not finish the step. When expectations were wrong, the fix may belong in marketing or signup messaging rather than in onboarding at all.

Start with the step that loses the most users relative to its importance. Resist the temptation to fix whatever is easiest. A copy tweak on a step that loses a few users is satisfying to ship, but it will not move activation the way addressing the main leak will.

Step 9: Measure the change

After shipping a fix, compare new cohorts with previous ones at the same step, using the same time window. Give it enough time to collect a reasonable number of signups, since small samples can make a fix look better or worse than it is.

Then check that the improvement carries through to activation. Sometimes fixing one step just moves the drop-off downstream. If more people now connect data but just as few reach activation, the next step has become the new bottleneck, or the people you rescued were never likely to activate. Either way, the funnel tells you what to look at next.

Common mistakes

The most damaging mistake is treating tracking gaps as user behavior, which sends a team to fix steps that are working fine. Another is looking only at overall conversion from signup to activation, which tells you there is a problem but not where. Ignoring segments hides cases where one channel or company size behaves very differently.

Teams also commonly fix the step that is easiest to change instead of the one losing the most users, and many never talk to the users who left. The interviews in step seven are the cheapest research you will ever do, and they regularly overturn assumptions.

Make it a routine

Review your onboarding funnel once a month. The biggest leak will change as you fix problems and as your audience evolves, for example when a new marketing channel starts bringing in a different kind of user. A short monthly review, using the same steps and time window each time, keeps onboarding improving steadily and spares you the large redesigns that come from ignoring the funnel for a year.