How to Investigate Onboarding Drop-Off

Choose whether the unit is an attempt, a person, or an account. Document valid alternate paths and match later completions to their starts. Exclude your test activity. GOV.UK's completion-rate guidance emphasizes defined start and end points and handling journeys that resume later.

For a fictional account-level funnel, 100 eligible accounts begin setup and 60 reach the defined outcome within seven days. Completion is 60%, and 40% have not reached that outcome within the window. Those 40 accounts are an investigation group, not 40 confirmed dissatisfied customers. Give every account the full seven days before comparing cohorts.

Source: GOV.UK: Measuring completion rate

Read relevant support conversations and invite a mix of customers who stopped and customers who completed the task. Ask about their last attempt. Compare role, setup route, device, and prerequisites when those differences are relevant, without dividing a small dataset into groups too small to interpret.

Suppose three fictional customers stopped at data connection: one lacked administrator access, one was waiting for a colleague, and one could not understand an error. A new tooltip might help the third person while doing nothing for the others. Record each explanation with its evidence and what would contradict it.

Choose a response that matches the obstacle. A permission prerequisite may need earlier explanation and a clear handoff to an administrator. An ambiguous error may need a specific recovery instruction. Verify the proposed fix with the original task before extending it to the whole onboarding journey.

Define the outcome and observation window before release. Compare equivalent cohorts and note changes in acquisition, product access, or assisted setup. A before-and-after increase alone does not prove that your change caused it. If a causal estimate matters, use an appropriately designed experiment and keep the customer evidence alongside the results.