Overview
Where we came in
A dental group with six practices, taking new patient bookings through a form on each practice page. This was an administrative problem, not a clinical one: the site was asking people to self-select a treatment type before they had spoken to anyone. We changed how the appointment is requested. We did not touch anything about care itself.
The baseline problem
- The appointment type dropdown listed 11 clinical options with no descriptions.
- Form analytics showed the dropdown was the last interaction for 34% of people who abandoned the form.
- Reception staff reported that a large share of submitted bookings had the wrong appointment type and had to be corrected by phone anyway.
- The form asked for date of birth and address before it asked what the person actually wanted.
Hypothesis
What we expected, and why
Written before the test was built. Because we observed X, we expect Y, measured by Z.
Because first-time patients cannot reliably choose between clinical appointment categories, we expect that replacing the dropdown with plain-language reasons for visiting will increase completed booking requests and reduce the number of requests reception has to reclassify.
Experiment plan
How the test was set up
Sample size and stop date were both fixed before launch, so no result here was called early.
Two-arm split test on the booking form across all six practice pages. Control kept the 11-option clinical dropdown. Variant offered four plain-language options covering the reasons people gave in their own words, with an open text field, and moved date of birth and address to a confirmation step after the request was submitted.
- Primary metric
- Completed booking requests per practice page visitor
- Traffic
- ~14,900 practice page visitors per arm
- Duration
- 30 days across all six sites
- Guardrails
- Reception reclassification rate, the operational cost of getting this wrong
- Requests missing information needed to confirm the booking
- Appointment attendance rate
Results
What actually moved
| Metric | Control | Variant | Change |
|---|---|---|---|
| Completed booking requests | 12.28% | 15.66% | +27.5% |
| Form abandonment | 44.1% | 29.8% | -14.3pp |
| Requests reclassified by reception | 38.2% | 16.4% | -21.8pp |
| Requests missing required details | 6.1% | 6.4% | +0.3pp |
| Appointment attendance rate | 87.9% | 88.3% | +0.4pp |
The reclassification number is the one the practice managers cared about. Fewer corrective phone calls freed reception time, and that operational saving was larger than the booking uplift in the first quarter. Attendance was flat, which is what we wanted: this test was about the request, not about care.
What changed
Shipped to production
- The 11-option clinical dropdown was replaced with four plain-language reasons for visiting, plus an open text field.
- Date of birth and address moved to a confirmation step after the request is submitted.
- Each practice page now states its opening hours and next available slot near the form.
- A visible note explains what happens after submission and when to expect a call back.
Next experiments
Already in the queue
- 01Test showing real available slots inline instead of a request-and-wait model.
- 02Test an SMS confirmation against the current email-only acknowledgement.
- 03Test the plain-language pattern on the existing-patient rebooking flow.