Onboarding Metrics
Surfer — Standard Plan
v1.2 — May 2026
Krystian Prorok
TARGETHard number — failing this triggers actionSIGNALDiagnostic only — watch, don't optimize yet
Primary metrics
> 90%
Completion rate
Paid users with clear intent — below 90% signals a broken step, not a funnel problem.
< 5 min
Time to complete
Is the flow fast enough to not lose people?
> 50%
Activation rate
% who open their first visibility report within 48h. Proves the onboarding promise was kept.
Per-step metrics
| Metric | What it tells you | Target / Signal |
|---|
| GSC connect rate | Are users comfortable connecting Google accounts? The only real exit risk in the entire flow. | > 60% |
| Edit rate per AI-generated field | Which fields does the AI get wrong? | <10% = not engaging20–40% = healthy>60% = AI is wrong |
| Time on step | Is reviewing AI suggestions causing friction? | < 90s |
| Drop-off rate | Paid users rarely abandon mid-setup — a spike here means something broke (OAuth error, missing GSC access), not a funnel problem. | Diagnostic only |
| Metric | What it tells you | Target / Signal |
|---|
| Average changes (adds + removes) | How accurate is the AI pre-selection? Under 5 changes = working well. | < 5 changes |
| Time on step | Users review 50 articles — scanning takes time. Under 120s is efficient. | < 120s |
| Drop-off rate | Should be near zero — pre-filled, low effort. Any spike is a bug, not user intent. | Diagnostic only |
| Metric | What it tells you | Target / Signal |
|---|
| Prompt deviation score | How far is the final selection from what AI proposed? | Baseline only |
| Change pattern (added vs removed) | What types of prompts did users add or remove? Clusters of similar changes reveal which topics the AI consistently gets wrong. | Baseline only |
| Time on step | Category review needs a real read — under 90s is efficient without being rushed. | < 90s |
| Drop-off rate | Should be near zero — low effort, skippable in feel. Any spike is a bug. | Diagnostic only |
3.5
Team Invite optional · skippable| Metric | What it tells you | Target / Signal |
|---|
| Team invite rate | Inviting here means the first AI visibility report lands in the whole team's inbox together — shared context drives retention. | Baseline only |
| Skip rate | High skip rate is expected — solo users will always skip. If >80% skip → move post-onboarding. Easily reversible decision. | Watch |
| Metric | What it tells you | Target / Signal |
|---|
| Return visit < 24h | Did the user come back to see first ChatGPT data? The onboarding promise is only kept if they return. | > 50% |
| Time to first data | How fast does the product deliver value? Directly tied to the return visit rate above. | < 24h |
Post-completion engagement
Onboarding drop-off should be near zero for paid users. This table tracks whether users return to see their first AI visibility report — the real proof that the product delivered on its promise.
| Segment | Definition | Signal | Action |
|---|
| Same-day | Returns < 24h | High intent — onboarding promise landed | None needed |
| Nudged | Returns after email nudge | Nudge is working | Optimise nudge timing |
| Organic (2–7 days) | Returns without nudge | Recoverable | Send nudge before day 7 |
| Churned | No return after 7 days | Lost — product didn't deliver perceived value | Diagnose: data delay? Wrong expectations set? |
Key lens: Split all metrics by connection method (GSC vs URL). URL users are expected to show lower pre-selection accuracy, longer time on step 2, and higher drop-off. That split is the primary diagnostic for what to improve first.