Most teams treat testimonials as decoration: pick a nice quote, drop it on the page, and never touch it again. But a testimonial is a conversion element, not wallpaper. The specific quote you show, where it sits relative to the call to action, and whether it carries a photo, a name, and a result all change how many visitors trust you enough to act. A/B testing is the only way to replace guesses with evidence — but it is also easy to run badly, waste weeks, and walk away with a "winner" that was really just noise. The honest answer to "should you A/B test testimonials?" is: yes, once you have enough traffic to get a clean answer, and only after you have fixed the obvious problems that don't need a test at all.
When testing is worth it — and when it isn't
A/B test needs traffic. To detect a realistic lift — say a testimonial change that moves conversion from 4% to 4.6% — you typically need thousands of visitors per variant over one to two full weeks. If your landing page gets a few hundred visitors a month, a test will run for a quarter before it reaches significance, and by then the market, your product, and your traffic mix have all shifted. For low-traffic pages, skip testing and make decisions from principles instead: put your strongest, most specific testimonial near the call to action, add a real name and photo, and move on. That advice is covered in where to place your strongest testimonial on a landing page.
Testing earns its keep when three things are true: you have real traffic (roughly a few thousand conversions-worth of visitors per month), the page matters commercially, and you have already exhausted the changes that are obviously right. Don't burn a test slot proving that a photo beats no photo, or that a named customer beats "Anonymous" — those are settled questions, and you should just add the photo and the name. Reserve tests for the genuinely uncertain calls where reasonable people disagree.
What to test first
Not all testimonial variables are equally worth testing. Order them by expected impact:
- Which testimonial you show. The single biggest lever. A quote that names a specific, quantified result ("cut our reporting time from a day to twenty minutes") almost always outperforms a warm-but-vague one ("the team is lovely to work with"). If you only run one test, test your two strongest candidate quotes against each other.
- Placement relative to the CTA. A testimonial the visitor reads just before they decide often beats one buried at the bottom. Test moving your best quote to sit immediately beside or above the primary button.
- Format: single hero quote vs. a cluster. One prominent, credible testimonial versus a grid of several shorter ones. The right answer depends on your audience — some are reassured by volume, others by a single deeply relevant story. This is a real toss-up worth a test, discussed in should you use a testimonial slider or a static grid.
- Framing elements. Star ratings, company logos, or a one-line result headline pulled from the quote. These are smaller levers — test them only after the big three.
Resist the urge to change several things at once. If you swap the quote and move it and add a logo in the same variant, a win tells you nothing about which change earned it. Isolate one variable per test.
Running a test that produces a real answer
The most common way testimonial tests go wrong is calling a winner too early. Conversion rates swing wildly in the first few days — a variant can look 40% ahead on Tuesday and be dead even by Friday. Decide in advance how long the test will run and how many conversions each variant needs before you look, and then don't peek-and-stop the moment one variant is ahead. A pre-registered sample size and a fixed end date protect you from your own optimism.
Run the test for at least one full business cycle — usually two weeks — so that weekday and weekend traffic, and any weekly email or ad rhythms, are represented in both variants equally. Make sure the only difference between A and B is the testimonial change itself; if the variant also loads slower or shifts the layout, you are testing that instead. And be honest about significance: a 3% "lift" with a p-value of 0.4 is not a result, it is noise wearing a result's clothes. If the test is inconclusive, that is a legitimate outcome — it means the two options are close enough that you can pick either on other grounds.
One practical note: keep a record of what you tested and what happened, even the flat results. Over a year, a log of "the quantified-result quote beat the friendly quote three times" becomes a reliable house rule you no longer have to test — which is the real payoff of testing in the first place. And whichever quote wins, make sure it is one you have written permission to use; see how to ask permission to use a customer's job title and company name in a testimonial.
The bottom line
A/B testing testimonials is worth it when you have the traffic to get a clean answer and have already made the obvious improvements. Test the highest-leverage variables first — which quote, and where it sits relative to the CTA — change one thing at a time, and let the test run its full course before declaring a winner. On low-traffic pages, skip the test and lean on principles: a specific, quantified, attributed testimonial placed near the decision point is a safe default that rarely loses. Testing is how you refine that default once you have enough visitors to hear the signal over the noise.