Dental website conversion rate optimisation: judged on booked consultations, not form fills

Where dental enquiries are lost

A patient who books from your website passes through a chain of small steps. They land, find the page for their treatment, find the answers they need, pick a route (phone, form or online booking), complete it, and then someone at the practice turns that enquiry into a time in the diary. Each step can leak.

These are the leaks I check first, and how each one is measured.

The last two rows are not website faults, but they are where website enquiries die, so I measure them in the same review. Missed calls have their own page on what missed calls cost a practice. Watching what visitors do on the page itself, through analytics paths and recordings, is covered in reading how patients use your website. Finding these leaks is one of the first jobs, after the research, in what an engagement includes.

Measure booked consultations, not form fills

Measure booked consultations because a form fill is not a patient. Forms collect spam, existing patients asking about appointments, people looking for an NHS dentist and job applicants. A tap on a phone number records the tap, not whether the call connected or anyone answered it. If you judge the website on these, a change that brings more forms but fewer bookings looks like a win.

This is the same reasoning that replaces cost per lead with cost per booked consultation. If ads send traffic to the site, Google lets you mark conversion actions as primary, used for bidding and reporting, or secondary, recorded in "All conversions" but not used for bidding.² I make the booked consultation primary and keep taps and forms secondary. A conversion action is the thing Google is told counts as a result; other terms are in the dental marketing glossary.

Joining website enquiries to bookings needs two things: each route tracked with its source, and reception recording that source against the new patient. The tracking side is in call tracking for practices and tracking online bookings.

Fix first, then test

A fix is a change to something broken or missing, where nobody could sensibly argue for the old version. A test is for a change where a reasonable person could argue either way. A/B testing means showing two versions of a page to visitors split at random and comparing what each group does. Testing a fix wastes traffic you could have spent on a real question.

The booking flow, phone and forms

The booking flow is every step from "I'll get in touch" to a time in the diary. I walk it on a real phone, on mobile data, for every route, before I change anything.

The phone. The number sits in the header on every page as a tappable link, with opening hours next to it, and says what happens out of hours. If you use call tracking, the tracked number must be the one patients tap, or the calls go unattributed.

Forms. Ask for what reception needs to book: name, phone or email, the treatment, and a preferred time. Leave out medical history. UK data protection law requires personal data to be "adequate, relevant and limited to what is necessary"⁴, and health details sent through a website form are not needed to book a consultation. After submission, the confirmation page says what happens next and when.

Online booking. Show the consultation fee and what it includes before the patient picks a slot. If the booking tool runs on another company's domain, check that the source survives the hand-off, because that is where tracking most often breaks.

Trust signals. These are the things that tell a patient the practice is real and competent: clinicians named with their GDC numbers, genuine reviews, photos of the actual practice, clear prices. Several are also legal requirements, set out in the details a UK dental website has to show.

Speed. Google's measures of a good experience are a Largest Contentful Paint within 2.5 seconds, an Interaction to Next Paint of 200 milliseconds or less and a Cumulative Layout Shift of 0.1 or less, at the 75th percentile of page loads.¹ Large images and third-party widgets are the usual culprits.

Two neighbouring jobs have their own pages. The page a patient reaches from an ad is covered in how I build pages for ad traffic, and the structure of a treatment page in what a treatment page needs.

Persuasion the rules do not allow

General CRO advice borrows from online retail: countdown timers, "only two left", pressure to decide now. Much of that is unlawful or against the advertising rules for anyone, and it sits badly with a dentist's professional duties. This is general information, not legal advice.

The banned practices in Schedule 20 have applied since 6 April 2025.⁵ A real deadline, stated accurately, is allowed. The rules target false urgency and false scarcity. More detail on reviews is in the rules on reviews and testimonials.

When there is too little traffic to test

Most single practices don't get enough visitors to run a clean A/B test on booked consultations. Government guidance on A/B testing is blunt about it: "you will need many users for the data to be statistically significant."¹⁰ Judging on bookings rather than form fills makes this harder, because bookings are rarer. I would rather say that plainly than run a test that cannot give an answer.

The table below shows the arithmetic. Illustrative figures, not benchmarks: I chose a starting rate of 3 bookings per 100 visitors to make the maths readable, and worked the sample sizes with the standard two-proportion formula at 5% significance and 80% power.

Small improvements need far more traffic than most practices have. So at low traffic, I work like this:

1. Fix the leaks. Fixes don't need a test, and they are usually where the lost bookings are.
2. Test only big differences. A different booking route might show up in your numbers. A different button colour won't.
3. Make one change at a time and compare equal periods. Compare booked consultations from the website for the weeks before and after, and note anything else that moved: ad spend, school holidays, a new clinician. Treat the result as an observation, not proof.
4. Use direct evidence. Book through every route on your own phone. Ask reception what callers ask that the website should have answered.
5. Decide the length of any test before it starts, and don't stop it early because one version is ahead.

If you do run a split test, Google asks you not to show search engines a different version from visitors, to point test URLs at the original with a canonical tag, to use temporary 302 redirects, and to remove the test once it has concluded.¹¹

If you want to know where your site loses enquiries, email me at [email protected] with your web address. The first look is free: I go through your site from the outside, as a patient would, and reply with what I find. In an engagement, I start by running a test enquiry through every route and following it into the diary. The rest of the process is on how I work.