Rebuilding Supanote's PLG activation path
Signup→subscription sat flat at ~12% for months. I reframed the quarter around conversion, found the real leak — setup, not the demo note — and lifted trial-to-paid +5% in two months.
Context
Supanote is an AI clinical scribe for behavioral-health providers — therapists, psychiatrists, counsellors. Acquisition was healthy: Facebook ads, Reddit, and the primary volume from ICP-targeted blog content were landing the right users on the signup page. But signup→subscription conversion sat at roughly 12%, flat for months despite a growing top of funnel.
The problem
The team’s reflex was to keep shipping features. I argued the opposite: the constraint wasn’t acquisition or surface area — it was everything that happened after signup. So I made signup conversion the theme for the quarter and moved the team off feature work and onto the funnel.
More traffic into a leaky middle doesn’t compound — it just costs more.
Approach
Define activation first. Before optimizing anything, I needed a target that actually correlated with revenue. One threshold stood out: users who created 5 notes within 7 days almost always subscribed — the 5th-note and subscription numbers tracked so closely they were effectively the same event. That became the activation metric: the leading indicator, distinct from the lagging signup→subscription outcome.
Instrument the whole path. I instrumented the funnel end to end against the 7-day free trial (in PostHog + GA4, with SQL on top) so every stage had a number and a drop attached: Signup → Setup complete → First note (24h) → 5th note / activation (7 days) → Subscription.
Diagnose — sessions vs. funnel. PostHog replays were emphatic: over 80% of sessions ended right after users saw the demo note, pointing to “the demo note is the drop-off.” But the full analytics funnel showed the bigger, more fixable leak upstream: setup completion was shedding 20–47% of users before they ever reached a real note. The replays showed a symptom; the funnel showed the cause.
Sessions tell you where it hurts. The funnel tells you where to operate.
Ship.
- Rebuilt the setup flow — reworked the highest-friction screens and removed the steps causing the 20–47% drop, collapsing the path from signup to a provider’s first real note.
- Pulled the activation moment forward — reoriented onboarding around first note in 24 hours and 5th in 7 days, inside the trial window.
- Shipped reliability fixes in parallel — production incidents were quietly taxing trust during the exact window users were deciding whether to pay.
Impact
- +5% trial-to-paid conversion lift in two months, off a flat ~12% baseline.
- Drove the production incident rate from 3.5% to under 1% in parallel.
Reflection & tradeoffs
- Activation and conversion are different layers — say so out loud. Signup→subscription is the lagging business outcome; activation (5 notes in 7 days) is the leading indicator you actually move to get there. Keeping them distinct kept the team honest about what each change was supposed to do.
- Qualitative finds the question; quantitative answers it. The replays were essential — they told me where to look — but committing to the demo-note story would have been the wrong bet. The funnel redirected the quarter to the leak that actually mattered.
- What’s next: the part trials don’t show you — retention cohorts past the trial (D7/D30) and a referral loop into a network of providers who already trust the tool.