Pricing page A/B tests: 12 experiments worth running
The pricing page is where visitors do arithmetic about you. They compare tiers, guess which limits they will hit, look for the catch, and decide whether the number feels fair. Every element on the page is an input to that arithmetic — which makes it dense with testable decisions, and also easy to test badly, because the metric that matters (revenue) is further from the click than on any other page.
Here are twelve experiments grouped by the decision they target, followed by the measurement pitfalls that make pricing tests lie.
Anchoring and tier structure
- Tier order: expensive-first vs cheap-first. Leading with the top tier anchors the rest as reasonable; leading with the cheap tier lowers the entry threshold. Which effect dominates depends on your audience's price sensitivity — a genuinely open question worth settling.
- Highlighted tier. Move the "most popular" badge and visual emphasis between tiers. This is among the strongest levers on the page because defaults are powerful, and it changes revenue per signup, not just signups.
- Number of tiers shown. Test three visible tiers against four, folding the extra into a "talk to us" link. Fewer comparisons can mean faster decisions; it can also remove the decoy that made your middle tier look good.
- Enterprise anchor row. Adding a "custom" column with no price anchors the paid tiers as the affordable options. Cheap to test, occasionally meaningful.
The annual toggle
- Default state: annual vs monthly. Defaulting to annual shows lower per-month numbers and shifts the mix toward annual plans — but the displayed price is only real if they commit for a year, and visitors notice. Measure plan mix and refund/churn behavior, not just clicks.
- Discount framing. "Two months free" versus "save 17%" versus showing the struck-through monthly total. Same economics, different comprehension; the winner is an empirical fact about your audience.
Information depth
- Feature table depth. A full comparison grid versus five headline features per tier with an expandable "see all". Exhaustive tables serve committed evaluators; short lists serve everyone else. Segment results by traffic source if you can.
- Plain-language limits. Rewrite quota rows ("50k MTUs") as consequences ("enough for ~50,000 monthly visitors"). Comprehension experiments are undervalued because they are unglamorous.
- FAQ placement. Pricing FAQs at the bottom versus inline beside the tiers versus expandable under each plan. The FAQ answers objections; the test is whether answering them earlier beats keeping the page scannable.
Reassurance
- Proof near the decision. A short customer quote or usage line adjacent to the CTA, versus a logo strip, versus nothing. Placement usually matters more than volume — see social proof experiments for why specificity wins.
- Risk reversal. Test the prominence of the money-back guarantee, free tier, or "cancel anytime" line near the buttons. Only claim what is true; the test is about visibility, not invention.
- CTA copy per tier. "Start free trial" versus "Get started" versus value-forward labels. Small lever, but it compounds with everything above; the mechanics are covered in CTA copy experiments.
How pricing tests lie to you
Pricing pages have the widest gap on your site between the click and the money, and every step of that gap is a chance to fool yourself.
- Signups are not revenue. A cheaper-feeling page reliably lifts signups while shifting the mix toward lower tiers. If you stop measuring at the click, you will ship variants that shrink revenue. Measure revenue per visitor, or at minimum plan mix.
- Low volume, long confirmation. Purchases are rarer than clicks and trial conversions land weeks later. Pricing tests therefore need more patience than headline tests; sequential statistics help because you can watch continuously and stop the moment evidence is genuinely sufficient, but they do not make sparse data dense.
- Seasonality and sales motions. A quarter-end discount push or a launch spike changes who visits the page. Randomization protects the comparison, but interpretability still benefits from running through a full cycle.
- Existing customers see it too. Plan changes and upgrade flows often route through pricing. Decide upfront whether current customers are in the test population, and be careful about changing displayed prices for them mid-cycle.
Sequencing the twelve
Do not run these as a redesign. Start with the structural levers — highlighted tier, tier order, annual default — because they move the most money and their results change how you interpret everything else. Then information depth, then reassurance polish. One experiment at a time on the same page keeps the reads clean.
Mechanically, each test is a small code change with disproportionate care required: assignment, price display logic, and cleanup on a page where bugs cost money. That is the part Trevo does for you — it proposes the next pricing experiment from your funnel evidence, writes the variant as a pull request your team reviews before anything ships, and runs the analysis with always-valid statistics. The judgment about what your product costs stays yours.
Frequently asked questions
What should I measure in a pricing page A/B test?
Pick one primary metric before launch: revenue per visitor is the honest default, with paid conversions acceptable when volume is too low. Track plan mix and downstream churn or refunds as secondary metrics, because pricing variants routinely trade signup volume against revenue per signup, and clicks alone cannot see that trade.
Do pricing page tests need more traffic than other tests?
Usually yes. Purchases are much rarer than clicks, so the same relative lift takes far longer to detect on a revenue metric than on a click metric. Teams with modest traffic should test bigger structural changes — tier order, highlighted tier, annual default — where plausible effects are large enough to detect in reasonable time.
Is it OK to A/B test different prices?
Testing presentation (order, emphasis, framing) is uncontroversial. Testing actual price points is heavier: customers can discover they paid different amounts, which is a trust risk, and in some contexts a legal question worth checking. Many teams instead test prices sequentially across cohorts or geographies, accepting weaker inference for cleaner customer treatment.
Should the annual or monthly price be the default toggle?
Annual-default shows lower per-month numbers and tilts the mix toward annual commitments; monthly-default reads as lower risk. Neither is universally right. Test it with revenue per visitor and plan mix as the metrics, and label annual pricing clearly — a visitor who feels tricked by "billed annually" fine print is more expensive than the lift is worth.