Pricing Experiments
Pricing experiments test how price is presented — anchors, comparison framing, bundle structure — rather than the headline number itself. Here's how to design them safely.
Pricing Experiments
Controlled tests of how a price is anchored, framed, or bundled — measured on revenue per visitor, not just conversion rate.
A pricing experiment is a structured test of a pricing variable — the anchor price, comparison frame, discount format, bundle composition, or currency presentation — measured against a control on a revenue metric rather than a UX metric. Unlike a typical A/B test on button copy, pricing tests sit inside a tighter operating envelope: showing two visitors materially different headline prices for the same SKU at the same time raises legal exposure under price-discrimination and consumer-protection rules in most EU and US jurisdictions, and a leak of that test to social channels can damage brand trust fast. The common workaround is to hold the headline price constant and test what surrounds it — the strikethrough anchor, the per-unit breakdown, the bundle default, or the urgency cue.
The reason pricing experiments deserve their own playbook — separate from regular conversion-rate tests — is that the downside is asymmetric. A bad button-copy test costs you a week of traffic. A bad pricing test can trigger refund demands, a chargeback spike, or a regulator letter. The design discipline exists to keep the upside (real revenue lift from better anchoring) without the legal and brand tail risk.
Pricing experiments sit at the intersection of pricing psychology — the cognitive shortcuts buyers actually use — and behavioral experimentation, the discipline of measuring those shortcuts in production. In practice, that means treating the test as you would any other split test: pre-registered hypothesis, sample size calculated upfront, single variable changed, and a guardrail metric (usually return rate or refund rate) running alongside the primary lift metric.
RPV_uplift = ((Revenue_variant / Visitors_variant) - (Revenue_control / Visitors_control)) / (Revenue_control / Visitors_control)
Revenue_variant
Variant revenue
Total order revenue from visitors in the variant arm over the test window.
Visitors_variant
Variant visitors
Unique visitors bucketed into the variant arm.
Revenue_control
Control revenue
Total order revenue from visitors in the control arm over the same window.
Visitors_control
Control visitors
Unique visitors bucketed into the control arm.
A Shopify apparel store tests a strikethrough anchor (€89 → €69) against a clean €69 price on its hero product page over two weeks.
Variant revenue: €42,400
Variant visitors: 12,000
Control revenue: €36,800
Control visitors: 12,000
→ +15.2% RPV uplift
Variant RPV is €3.53, control is €3.07. The strikethrough anchor lifted revenue per visitor by 15% — driven by a higher add-to-cart rate, not a price change. Worth shipping once you confirm the return rate guardrail held.
Use RPV — not conversion rate — as the primary metric. A pricing test can lift conversion while tanking average order value, or vice versa. RPV captures both effects in one number, which is what the P&L actually cares about. Pair it with a refund-rate guardrail watched daily; a variant that lifts RPV but ships a 4-point jump in returns is a loss disguised as a win.
Typical RPV uplift ranges for common pricing-experiment designs
| Test design | Apparel / accessories | Beauty / supplements | Electronics / home |
|---|---|---|---|
| Strikethrough anchor on PDP | +8% to +18% | +5% to +12% | +3% to +9% |
| Per-unit framing (e.g. €/day) | +2% to +6% | +6% to +14% | +1% to +4% |
| Bundle default swap | +10% to +22% | +8% to +18% | +5% to +12% |
| Charm pricing (.99 vs .00) | +1% to +4% | +1% to +3% | +0% to +2% |
| Free-shipping threshold change | +4% to +11% | +3% to +8% | +2% to +7% |
Two practical guardrails. First, segment by traffic source before you call a result — paid social traffic anchors very differently from branded search traffic, and a winning variant on one can be flat on the other. Second, run pricing tests for full weekly cycles (minimum 14 days), because weekend buyers and weekday buyers price-shop differently and a 7-day test almost always over-indexes on one or the other.
Pricing experiments: common questions
In the EU and UK, showing materially different headline prices for the same SKU to comparable visitors at the same time is risky under consumer-protection and unfair-commercial-practices rules, and personalised pricing based on profiling triggers GDPR transparency obligations. The standard workaround is to test the presentation around the price — anchor, framing, bundle — while holding the headline number constant.
It's the most-tested pricing variable and usually the lowest-impact one for stores above a €30 AOV. Run it once to settle the debate for your category, then move on to anchor and bundle tests where the uplift ranges are 3-5x larger.
Minimum 14 days to cover two full weekly cycles, and long enough to reach your pre-calculated sample size on RPV. Pricing tests need more sample than conversion tests because RPV has higher variance than a binary conversion event.
A promotion changes the price for everyone over a window; a pricing experiment splits traffic to measure the causal effect of a pricing variable. You can run promotions without a control arm — you can't learn from them the same way.
Yes, and it's often the safest design — pick 3-5 representative SKUs, run the test there, and roll out site-wide only if the lift replicates. This limits both blast radius and the chance of price-comparison screenshots circulating.
Pricing psychology is the body of theory — anchoring, decoy effects, left-digit bias, loss aversion. Pricing experiments are how you find out which of those effects actually move revenue in your store's specific context, rather than assuming the textbook result applies.
RPV (revenue per visitor). Conversion rate can move opposite to AOV in a pricing test — a lower-priced variant often converts more but earns less per visitor. RPV captures both in one metric and aligns with the P&L.
Refund rate, return rate, and customer-service ticket volume — all on a 7-day rolling basis. A variant that lifts RPV but drives a 3-point return-rate increase is usually a net loss once reverse-logistics costs are included.
Technically yes, but most teams run them through their CRO platform so the bucketing, sample-size math, and revenue attribution are handled in one place. Splitting pricing logic across a flag tool and an experimentation tool is where attribution bugs creep in.
Lower than UX tests — roughly 1 in 4 pricing experiments ships a meaningful winner, versus 1 in 3 for typical PDP or checkout tests. The wins tend to be larger, though: a successful anchor or bundle test commonly delivers 8-15% RPV lift, which is a category most other CRO work can't reach.
Test ideas before you ship them
Run unlimited A/B tests, attach hypotheses to outcomes, and build a searchable archive of what works — and what doesn't.