Ran multiple hypothesis-driven experiments across no-shows, cancellations, and booking flows
5 Experiments
Each experiment below followed the same discipline: observe something in the data, form a hypothesis, design a test, and measure the outcome. No gut feelings — just data-backed decisions.
No-show rates running at ~7%. Users were forgetting sessions when booking multiple at once.
Adding "Add to Calendar" functionality would reduce no-shows as a lightweight behavioural nudge.
Implemented calendar integration and measured no-show rates before/after.
Takeaway
Sometimes the simplest intervention has the biggest impact.
30% of users who cancelled cited "scheduling conflict" — they wanted a different time, not to quit.
Offering a reschedule-forward flow instead of a cancellation button would retain users who just need flexibility.
Replaced the straight cancellation flow with a "reschedule first" prompt.
Takeaway
Users were telling us why they cancelled — we just had to listen to the data.
Auto-matched users showed low commitment — 45% match-to-booking, 57.6% session completion.
Letting users pick their own provider creates psychological ownership and stronger commitment.
Cohort experiment replacing auto-match with user-selected provider listing.
Takeaway
User agency drives commitment. Auto-matching optimises for speed; user-selection optimises for outcome.
30% drop between matching with a provider and actually booking.
Combining match and booking into a single committed flow removes the friction gap.
Single-step flow where users commit to booking at point of provider selection.
Takeaway
Reducing steps reduces drop-off. Every extra click is a chance to lose the user.
10% of providers proactively sent session assignments (homework/exercises) to users.
Provider-initiated engagement between sessions improves retention.
Users with assigned providers had 12% higher retention.
Recommendation
Surfaced as a product recommendation to expand assignment adoption platform-wide.
Takeaway
The best insights sometimes come from observing outlier provider behaviour.
Looking for a data person who can go from SQL to boardroom? I'd love to chat.