RPV vs Conversion Rate
Conversion rate and revenue per visitor often disagree on the winner. Here's when each metric is the right one to call an A/B test — and when CR will mislead you.
RPV vs Conversion Rate
RPV measures revenue per visitor (CR × AOV); conversion rate measures only the share of visitors who buy — so they can disagree on the winner.
Conversion rate (CR) tells you what fraction of visitors completed a purchase. Revenue per visitor (RPV) tells you how much money each visitor generated on average — mathematically, CR multiplied by average order value (AOV). The two metrics agree most of the time, but they diverge whenever a change affects basket size: a test that adds a free-shipping threshold, reorders product tiles, or surfaces a higher-priced bundle can lift RPV while flattening or even depressing CR.
For most A/B tests on a Shopify or WooCommerce store, RPV is the safer decision metric because it captures both halves of the revenue equation. CR is still useful — it's faster to reach significance and clearer to diagnose — but calling a winner on CR alone risks shipping a variant that converts more shoppers into smaller orders.
The short version: if your test could plausibly change what people buy — not just whether they buy — read it on RPV. If the test only affects friction on a single product or flow (a checkout field, a CTA color, a form layout), conversion rate is usually enough.
The trap is that CR moves faster and reaches significance sooner, so teams default to it. That works fine for top-of-funnel and checkout tests. It breaks the moment your variant touches merchandising, pricing, bundling, or shipping rules — anywhere AOV is in play.
Same A/B test, two metrics, opposite winners: a free-shipping threshold raised from €50 to €75 on an apparel store.
| Metric | Control (€50 threshold) | Variant (€75 threshold) | Lift | Winner? |
|---|---|---|---|---|
| Conversion rate | 3.2% | 2.9% | -9.4% | Control |
| Average order value | €62 | €78 | +25.8% | Variant |
| Revenue per visitor | €1.98 | €2.26 | +14.1% | Variant |
| Visitors (per variant) | 48,000 | 48,000 | — | — |
| Total revenue | €95,232 | €108,864 | +€13,632 | Variant |
Calling this test on conversion rate would have killed €13,600 of incremental revenue over the test window. The variant lost some price-sensitive buyers but pulled the remaining ones into bigger baskets — a classic RPV-positive, CR-negative pattern.
When RPV is the better A/B test decision metric
Use RPV as the primary metric whenever your variant can change basket composition. That includes shipping-threshold tests, bundle and add-on placements, cross-sell and upsell modules, product-tile reordering, default variant selection (color, size, pack count), and pricing or discount-presentation changes.
RPV also wins for traffic-source comparisons. A Meta campaign at 4% CR but €38 AOV is worse than a Google campaign at 2.5% CR and €82 AOV — €1.52 RPV versus €2.05 RPV. Read channels on RPV, not CR, or you'll over-invest in cheap traffic that converts well but doesn't spend. This connects directly to conversion rate as a standalone metric: it's necessary but not sufficient.
Watch the AOV outlier trap
RPV has higher variance than CR because a single €800 order in a sample of 10,000 visitors moves the average noticeably. Always check the order-value distribution of both variants — if one side has a fat tail of unusually large orders, your RPV lift may be noise. Winsorising (capping orders at the 99th percentile) or reading the median order value alongside the mean usually settles the question.
When conversion rate is enough
CR is the right call when your test cannot plausibly affect AOV. Lead capture forms, newsletter signups, account creation, and checkout-field reductions all fit here — the conversion event is binary and the order value is set elsewhere. Single-product landing pages with one SKU and no upsells are CR-readable too.
CR also has a real practical advantage: it reaches statistical significance roughly 2-4x faster than RPV at the same sample size, because revenue is a noisier metric than a clean binary outcome. On a low-traffic store running tests against RPV fundamentals, that speed difference can be the difference between shipping six tests a quarter and shipping two.
Channel comparison: CR and RPV disagree on which traffic source is best
Conversion rate (%)
RPV (€)
RPV vs conversion rate: common questions
Conversion rate is the share of visitors who buy. RPV (revenue per visitor) is the average revenue each visitor generates — mathematically, CR multiplied by AOV. CR ignores order size; RPV captures both how often people buy and how much they spend.
Default to RPV whenever the test could affect basket size — merchandising, shipping thresholds, bundles, pricing, upsells. Use CR when the test only touches a binary friction point like a form field or a single-product CTA where AOV cannot move.
Your variant changed what people put in the basket, not just whether they checked out. Common causes: a higher free-shipping threshold, a more prominent bundle, a default to a larger pack size. Fewer customers, bigger baskets — net revenue up.
No. RPV is noisier than CR, so it needs more traffic to reach significance. For low-traffic stores or tests that genuinely cannot affect AOV, CR is faster and just as accurate. Match the metric to what the test can actually move.
RPV = total revenue ÷ total visitors. Equivalently, RPV = conversion rate × average order value. Both formulas give the same number; the second is more useful when diagnosing why RPV changed between variants.
GA4 doesn't show RPV as a standard column, but you can derive it: purchase revenue divided by sessions (or users, depending on your definition). Shopify Analytics shows AOV and conversion rate separately — you multiply them. Most CRO platforms surface RPV directly on test reports.
Yes — when a small number of unusually large orders skew the average. A single €1,500 wholesale-style order in a sample of 5,000 visitors can flip the result. Check the order-value distribution, winsorise outliers, or read median order value alongside RPV to sanity-check the lift.
By RPV. A channel that converts at 5% but with €30 AOV produces less revenue per visitor than one converting at 2% with €120 AOV. CR alone over-rewards cheap, low-intent traffic and under-rewards channels that bring high-value buyers.
Roughly 2-4x what you'd need for the same effect size on CR, because revenue variance is higher than binary outcome variance. As a rule of thumb, if you can detect a 10% CR lift in 14 days, expect 4-6 weeks to detect a comparable RPV lift on the same traffic.
Yes — always. Read RPV as the primary decision metric for tests that can affect AOV, but watch CR as a diagnostic. If RPV is up and CR is down, you know the win came from basket size; if both are up, you know the variant works for everyone.
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.