RPV vs Conversion Rate
A variant can lift conversion rate and still cost you money. Here's when revenue per visitor should arbitrate your A/B tests — and when conversion rate is the right call.
RPV vs Conversion Rate
Conversion rate measures how often visitors buy; RPV (revenue per visitor) measures how much they're worth — and the two often disagree.
Conversion rate (CR) and revenue per visitor (RPV) answer different questions about the same traffic. CR tells you what share of visitors completed a purchase. RPV tells you the average revenue each visitor generated, whether they bought or not — combining CR with average order value in a single number.
The two metrics agree most of the time, which is why teams default to CR. But they diverge whenever a variant shifts the AOV mix: discount banners, bundle prompts, free-shipping thresholds, upsell modals, and price anchoring all move buyers between baskets, not just into them. On those tests, picking the wrong primary metric ships the wrong winner.
The mechanical difference is one term. Conversion rate is orders divided by sessions. RPV is revenue divided by sessions — which is mathematically CR multiplied by average order value. Any test that touches pricing, discounting, bundling, or product mix moves AOV, and the moment AOV moves, the two metrics can point in opposite directions.
That's why most experienced CRO teams ship RPV as the decision metric for checkout, PDP, and pricing tests, and reserve conversion rate for top-of-funnel work where AOV is held constant. The pattern below is the one that catches teams out: a variant with a clear CR lift, a small AOV drop, and a flat or negative RPV.
Apparel A/B test: 10% off banner vs control (12,000 sessions per variant)
| Variant | Sessions | Conversion rate | AOV | RPV | Revenue |
|---|---|---|---|---|---|
| Control (no banner) | 12,000 | 2.40% | €78 | €1.87 | €22,464 |
| Variant (10% off banner) | 12,000 | 2.95% | €61 | €1.80 | €21,594 |
| Difference | — | +23% | -€17 | -€0.07 | -€870 |
The discount banner won conversion rate by 23% with a comfortable p-value. It also lost €870 over the test window because the average basket shrank faster than orders grew. If the team had set CR as the primary metric, the variant ships and quietly drains margin for the next quarter.
When conversion rate is the right call
Conversion rate is the cleaner metric whenever you're confident AOV won't move. That covers most pure-UX work: navigation changes, image quality, page speed, form field reduction, error-state copy, mobile menu redesigns, and trust signal placement. The variant changes how easily people complete an existing purchase intent, not what they buy.
CR is also the right metric for high-funnel tests where revenue is too noisy to read. Category page sort order, search relevance, and PLP filter UX all influence add-to-cart rate long before they influence basket size. Use CR (or micro-conversions like add-to-cart rate) as the primary metric and watch RPV as a guardrail.
The discount trap
Any variant that introduces, enlarges, or repositions a discount will almost always lift conversion rate. That's not a finding — it's the mechanic. If your test ships a price cut and you measure on CR, you've designed a test you can't lose and a P&L you can't defend. Use RPV (ideally net of discount) or margin-per-visitor.
When RPV must arbitrate
Switch to RPV the moment a test can move AOV. That includes free-shipping thresholds, bundle offers, quantity discounts, cross-sell modules, upsell modals on the cart, price display changes, and any variant that nudges product mix (featured SKU rotation, sort-by-price defaults, "trending" badges). On these tests, conversion rate is informative but not decisive.
RPV is also the right call for checkout-flow tests on stores with wide basket-size variance. A one-step checkout might convert lower-intent visitors who add a single low-margin SKU, lifting CR while RPV stagnates. Reading both metrics — and segmenting by basket size — tells you whether the variant grew the pie or just shuffled the slices. For the underlying mechanics, see RPV drivers.
Five common DTC tests: CR lift vs RPV lift
Conversion rate lift
RPV lift
RPV vs conversion rate: common questions
Use RPV when the variant can plausibly move average order value — discounts, shipping thresholds, bundles, upsells, price displays. Use conversion rate for pure UX, speed, and trust changes where AOV is held constant. When in doubt, ship RPV as primary and CR as secondary.
Because conversion rate ignores basket size. A discount or aggressive shipping offer pulls in price-sensitive buyers with smaller carts, lifting CR while dragging AOV down faster. Multiply the two and revenue per visitor falls — even though more people checked out.
No. RPV has higher variance than conversion rate because revenue is a long-tailed distribution (one €500 order can move the average). On low-traffic tests or tests where AOV is structurally stable, conversion rate reaches significance faster and is the more efficient choice.
RPV equals conversion rate multiplied by average order value. So a store with a 2.5% conversion rate and €80 AOV has an RPV of €2.00. This identity is why the two metrics agree when AOV is constant and diverge when it isn't.
Yes, materially. RPV needs more traffic because revenue variance is higher than the binomial variance of CR. Expect roughly 1.5x to 3x the sample size to reach the same significance, depending on your AOV spread. Power calculations on RPV should use your actual revenue distribution, not just the mean.
Yes — always. Primary metric on top, the other as a guardrail directly beneath. The two-up view catches discount traps, AOV erosion, and upsell wins that a single-metric dashboard would hide. Add margin-per-visitor as a third line if your SKUs vary in margin.
Same comparison, different shorthand. CVR (conversion rate) is orders / sessions. RPV (revenue per visitor) is revenue / sessions, which equals CVR × AOV. CVR tells you whether people bought; RPV tells you what those purchases were worth.
Raising a free-shipping threshold typically lowers conversion rate slightly (some buyers abandon) but lifts AOV more (others add an item to qualify). Net effect on RPV is usually positive, which is why this test should be judged on RPV — judging on CR would kill a profitable change.
Yes — for example when a variant skews toward one high-AOV SKU that's actually loss-leading or out-of-stock prone. Pair RPV with margin-per-visitor or contribution-per-visitor on margin-sensitive catalogues. RPV is the better default; it's not infallible.
Net is more honest for decision-making. Gross RPV flatters any discount-led variant. If your platform makes net revenue hard to read in-test, use gross during the experiment but recalculate net before shipping the variant — and reject any winner whose net RPV is flat or negative.
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.