Veterinary website A/B testing means showing two versions of the same page to real visitors and letting new-client inquiries decide the winner, instead of a team vote.
Done properly, it is the most honest instrument in marketing: visitors do not know which version they saw, and the only opinion that counts is what they book.
Done carelessly, it produces confident nonsense, and confident nonsense is worse than no data at all.
This guide is the mechanics chapter of veterinary website conversion, the parent guide to turning the traffic your site already gets into booked clients: what to test, how many visitors a valid test needs, and what to do when your traffic is too small for a clean one.
Split testing and A/B testing are the same thing
Split testing is the older name for the identical method: visitors are divided at random between the current page, called the control, and a challenger, everything else stays constant, and whichever version produces more of the goal wins.
The goal for a practice is never clicks or time on page; it is new-client inquiries per visitor, counting form fills, completed bookings, and phone calls that arrive from the site.
Multivariate testing, which changes several elements at once to see how they interact, is a different instrument, and it needs far more traffic than a clinic site carries.
A/B testing for vet clinics: what deserves a test
Testing effort should follow the leaks, because a month spent testing a page nobody visits is a month gone.
For most practices the candidates rank the same way: the new-client form first, the booking path second, the phone follow-up that catches everything the site starts, and the offers or pop-ups layered on top.
The form is the highest-intent moment on your site, and it has its own field-by-field guide in the veterinary new-client form.
The booking-versus-request-form question is a separate comparison, covered in veterinary online booking, and every rule below applies to it too.
Pop-ups can be tested as well, with one caveat from Google: the search team says intrusive interstitials and dialogs may lead to poor search performance and recommends banners instead, so if you test a pop-up, test it against a banner and watch your rankings while it runs.
Pick one change
The challenger differs from the control in exactly one way: a shorter form, a clearer booking button, an offer moved above the fold.
Write the hypothesis down first
One sentence before anything ships: we believe version B beats version A, measured by new-client inquiries per visitor, for a specific reason.
Fix the sample size in advance
Decide the visitor count before version B ever renders, using the arithmetic in the next section.
Run to the number, in whole weeks
Weeks capture the weekend dip every clinic site shows, and stopping mid-test corrupts the result.
Keep, discard, and queue the next
A winner stays, a loser is a cheap lesson, and the next test should already be chosen.
The sample size your traffic can support
Here is the constraint that keeps veterinary website A/B testing humble: valid tests are sized in visitors, and clinic sites do not carry many.
Work an illustrative example with a standard power calculation at 80% power and a 5% significance level, and say your site turns 5% of visitors into new-client inquiries.
Detecting a lift to 6%, a 20% relative improvement, takes roughly 8,000 visitors per version, about 16,000 in total.
Detecting a lift to 7.5%, a 50% relative improvement, takes roughly 1,500 per version, and doubling the rate to 10% takes roughly 450.
Visitors per version needed, from a 5% baseline
A clinic site with 1,500 visits a month would need most of a year for the small-lift test, about two months for the 50% lift, and under a month for a doubling.
Two lessons follow.
First, the subtler the change, the longer the test, which is why button-color experiments are a luxury for sites with traffic to burn.
Second, if a test would take a year, the honest move is to change the plan, not the arithmetic.
And never peek: checking a running test and stopping the moment it looks favorable can push the false-positive rate to 26.1% in Evan Miller's worked example, which is exactly why the sample size gets fixed before the test starts.
When traffic is too small: three honest options
Most single-location practices will not see 16,000 visitors in a quarter, and that is a reason to change the instrument, not to give up.
Option one is to test bigger swings: a challenger that changes the whole path, replacing the form instead of rewording a button, moves the rate enough to be detectable in far fewer visitors.
Option two is to run fewer, longer tests, each spanning whole months so a seasonal swing does not decide the winner for you.
Option three is before-and-after tracking: change exactly one thing, then compare the four weeks after against the four weeks before, accepting that it is weaker evidence than a controlled test because traffic drifts for reasons that have nothing to do with you.
Run tests like this
- One change per test.
- Sample size fixed in advance.
- Inquiries per visitor as the score.
- Whole weeks, start to finish.
Not like this
- Five changes at once.
- Stopping on the first favorable day.
- Clicks as the score.
- Calling it a win because you like version B.
Reading the result, and who runs this for you
There is no published veterinary conversion benchmark to measure a result against.
The closest sourced reference is Unbounce's 2024 report of 41,000+ landing pages: a 6.6% median across all industries and no veterinary row at all, so your own measured inquiry rate is both the baseline and the only comparison that matters.
For the full treatment of those numbers, see what a good veterinary website conversion rate looks like.
This is also the exact discipline behind the monthly CRO service described on the veterinary website conversion page: one controlled test a month on your site, with the result reported in plain English.
And it is a discipline with mileage on it: as Director of CRO at LaserAway from 2018 to 2023, Gabe took sitewide conversion from 3% to 11%, tested more than 2,600 variations, and returned 210x ROI on the testing program.
That record is med-spa booking conversion, not veterinary, and it is offered as experience that transfers, never as a client result.
If you would rather know your biggest leak before running anything, start with the free audit: a prioritized plan within 3 business days, no call required.
Frequently asked questions
How much traffic does a veterinary website need for A/B testing?
More than most clinic sites carry for small lifts. In a standard power calculation at 80% power, detecting an inquiry-rate lift from 5% to 6% takes roughly 8,000 visitors per version, while a 50% relative lift takes roughly 1,500 per version, which a busy site can reach in a month or two.
What is an example of an A/B test a vet clinic could run?
Show half your visitors the current new-client form and half a shorter version, then count new-client inquiries per visitor until the visitor count you fixed in advance is reached. One change per test, and the version that produces more inquiries stays.
How long should an A/B test run on a veterinary website?
Until the visitor count you fixed in advance is reached, in whole-week blocks so weekday and weekend traffic are both represented. Stopping the moment a test looks significant can make the result wrong: in Evan Miller's worked example, peeking pushed the false-positive rate to 26.1%.
Is A/B testing the same as split testing?
Yes. Split testing is the older name for the same method: visitors are split between two versions of a page, and the version that produces more of the goal wins. Multivariate testing is the different one, changing several elements at once and needing far more traffic than a clinic site has.
What should I measure in a veterinary website A/B test?
New-client inquiries per visitor, counting form fills, completed bookings, and phone calls that arrive from the site. Clicks and time on page are diagnostics, not the score.