Skip to content
SundAIFree call20 min
Menu
Browse all guides

200 people started signing up. Why did only 130 activate within a day?

In this worked example with fictional data, use PostHog to spot a gap, account records to check the count and Playwright to test a possible cause before redesigning the form.

Open the signup calculator

Start with a count you can defend

Suppose 200 distinct people started signing up and 130 activated an account within 24 hours. That is 65% completion. The other 70 had no recorded activation in that window. Some might return tomorrow. These numbers are a worked example. They say nothing yet about why anyone stopped. [6]

Count the same journey in two places

Give each person who enters the cohort a full 24 hours, including those who start on the last day. In PostHog choose an ordered funnel and count people once per step. Its overall and previous-step percentages use different denominators. Google Analytics documents a similar distinction between entering at the first step and entering midway. Match those settings before comparing reports. [1] [2]

  • Record three milestones: first interaction with the form, submission accepted by the server, and account activation. Record the outcome of each step without sending field contents.
  • If 160 of the 200 submitted, the first step retained 80%. Of those 160, 130 activated, or 81.25%. A report that shows 65% at activation has used the original 200 as its denominator.
  • Compare tracked activations with an aggregate from the account system for the same eligible group and time window. Check consent coverage, identity changes, retries and test accounts. Reconcile a mismatch before choosing a feature to change.

Turn one explanation into a browser test

Suppose support hears that people correct a misspelled email on a phone and then cannot continue. Give Playwright a test account in staging and the same screen size. Submit the invalid address, check that the page identifies the field, correct it and check that the confirmation appears. Link the visible result to one accepted submission in the test system. W3C’s form guidance provides a useful second check on the clarity of the error message. [3] [5]

Where Claude and MCP fit

A team can run that test directly in Playwright. If it already uses Claude with the Playwright MCP connector, an operator can ask it to walk through the same test and record the first difference between expected and observed behaviour. Keep the expected outcome in a repeatable test owned by the team. Use fictional details, a staging account and a test inbox. Keep customer recordings and production credentials outside the prompt. [4]

Decide what the next number would mean

Imagine that a later, comparable cohort records 150 activations from 200 starts. Completion has moved from 65% to 75%, a difference of 10 percentage points. A browser test can establish that the corrected address now works under the tested conditions. Changes in traffic mix or email delivery might also affect the observed increase. If a rollout decision requires causal evidence, plan a controlled comparison with an appropriate sample. [7]

What we can decide together

Bring two aggregate counts, the name of the step where customers get stuck and the outcome you expect them to see. In a free 20-minute call with Ariel, choose the first discrepancy to investigate. A scoped pilot can deliver an event map, one repeatable browser test, the result before and after a change, and a handover session with the person who owns the form.

How many people completed sign-up?

The starting figures are illustrative. Enter unique people from the same cohort, using one completion window and the same event definitions.

Sources

  1. PostHog · Funnels
  2. Google Analytics · Funnel exploration
  3. Microsoft Playwright · Best practices
  4. Microsoft · Playwright MCP
  5. W3C WAI · Form notifications
  6. GOV.UK Service Manual · Measuring completion rate
  7. GOV.UK Service Manual · Measuring the success of your service