How to use RPV for Landing Page Tests
Conversion rate alone hides the truth on landing page tests that touch price, bundles, or upsells. Here's how to use RPV as the primary win condition without losing decision speed.
RPV for Landing Page Tests
Using revenue per visitor as the primary win condition for paid landing-page experiments instead of conversion rate alone.
RPV (revenue per visitor) for landing page tests means scoring variants on the revenue each visitor generates, not just the percentage who convert. It's the right win condition whenever a variant can change average order value — different price anchors, bundle prominence, free-shipping thresholds, quantity selectors, or upsell modules all shift AOV in ways CR-only readouts miss.
For paid landing pages fed by Meta, Google, or TikTok traffic, RPV ties the experiment directly to the metric that determines whether the ad spend is profitable. A variant can lift conversion rate by 12% and still lose money if AOV drops 18%. RPV catches that; CR doesn't.
Most paid landing page teams still declare winners on conversion rate. It's the metric the testing tool surfaces first, it reaches significance faster, and it feels like the thing the landing page controls. None of that makes it the right scoreboard.
The problem is that landing page changes routinely move both conversion rate AND average order value — often in opposite directions. Switch the hero from a £49 single-product offer to a £79 bundle and CR usually drops while AOV climbs. CR-only readouts call this a loss. RPV tells you whether the till actually rang louder.
Why CR-only readouts mislead on landing pages
Three landing page changes reliably break the CR-as-truth assumption: price-perception shifts (anchors, strikethroughs, financing), bundle and quantity prominence, and upsell or cross-sell visibility above the fold. Each one can change AOV by 10-30% in either direction without you intending it.
A classic trap: you simplify the hero to push a single hero SKU and CR jumps 8%. But the previous version surfaced a 3-pack option that 22% of buyers were choosing. AOV collapses, and the cleaner page actually earns less revenue per visitor than the cluttered one you replaced.
The reverse trap is just as common. A bundle-forward variant pushes AOV from £52 to £71 but CR drops from 3.4% to 2.9%. On CR alone you kill the winner. RPV (£1.77 vs £2.06) shows the bundle variant is 16% more profitable per visitor — which on £40k/month of Meta spend is a meaningful number.
The CR-only failure mode
If your landing page test changes anything visible about price, bundles, quantities, or post-add upsells, a CR-only readout has a roughly 1-in-3 chance of picking the wrong winner. The bigger the AOV swing the variant induces, the higher that probability climbs.
Instrumenting RPV as the win condition
RPV is mechanically simple: total revenue from a variant divided by total visitors assigned to that variant. The instrumentation work is making sure the numerator captures the right revenue — first-order only, net of refunds and discounts, and attributed to the visitor's first assigned variant (not the variant they happened to be on at purchase, if you re-bucket).
On Shopify, the cleanest setup is to fire a purchase event with order subtotal (excluding shipping and tax) tied to the visitor's experiment ID. Subtotal — not grand total — keeps the metric comparable across regions where tax and shipping rules differ. For longer purchase windows, attribute revenue within a 7-day click window from first exposure.
Same test, two scoreboards: when CR and RPV disagree
Conversion rate (vs control)
RPV (vs control)
Four of the five variants above would be misjudged on CR alone. The simplified hero looks like a winner but loses 9% RPV. The bundle-forward variant looks like a loser but wins by 16%. This pattern — CR and RPV pointing in opposite directions — shows up in roughly a third of landing page tests once the variant touches AOV levers.
Benchmarks: what RPV typically looks like
RPV varies dramatically by vertical and traffic source, which is why an external benchmark is only useful as a sanity check. The number that matters is your own variant-vs-control delta. Still, knowing the rough range helps you spot instrumentation bugs (e.g. RPV reading £0.04 when it should be £1.80).
Meta-traffic landing pages tend to under-index on CR vs organic or email (colder intent) but can outperform on AOV when the LP is designed to bundle. Search traffic is the inverse: higher CR, lower AOV. RPV normalises this so you can compare LP performance across channels without arguing about which metric to weight.
Typical RPV ranges on paid landing pages by vertical and traffic source
| Vertical | Meta traffic RPV | Google Search RPV | Typical AOV | Typical CR |
|---|---|---|---|---|
| Apparel (€50-100 AOV) | €1.40 - €2.40 | €2.20 - €3.60 | €65 | 2.8% - 4.2% |
| Beauty (single SKU) | €0.90 - €1.80 | €1.60 - €2.80 | €38 | 2.5% - 3.8% |
| Beauty (subscription LP) | €2.10 - €4.50 | €3.40 - €6.20 | €52 | 3.5% - 5.5% |
| Home & accessories | €1.80 - €3.20 | €2.80 - €4.80 | €78 | 2.2% - 3.4% |
| Consumer electronics | €3.50 - €7.50 | €5.20 - €11.00 | €185 | 1.8% - 2.8% |
If your RPV sits well below these ranges with a healthy CR, the diagnosis is usually one of three things: discounting eroding subtotal, an upsell module that's been removed, or a checkout flow that's silently downselling buyers from the bundle they clicked on.
Worked example: Meta-traffic LP for a £79 apparel bundle
Picture an apparel store running £40k/month of Meta spend to a single landing page selling a £79 three-piece bundle. The control hero shows the bundle. The challenger drops to a single £39 hero piece to lift CR. Over 14 days: control sees 38,200 visitors, 1,184 orders (3.1% CR), £62 average subtotal, total £73,408 — RPV £1.92. Challenger sees 38,050 visitors, 1,560 orders (4.1% CR), £41 average subtotal, total £63,960 — RPV £1.68.
On CR the challenger wins by 32%. On RPV the control wins by 14%. Over a year of Meta spend at this volume, picking the CR winner costs roughly £110k in lost revenue — before you account for the lower-AOV cohort's worse repeat behaviour. This is the kind of decision RPV exists to get right, and it's also why broader landing page optimization work should default to RPV the moment a variant touches what's being sold, not just how it's framed.
Practical guardrails when you adopt RPV as the win condition: pre-register the primary metric (RPV) and the guardrail (CR floor — e.g. challenger must not drop CR more than 20% even if RPV wins). Run the test to revenue significance, not CR significance — revenue distributions are noisier, so you typically need 1.5-2x the sample size you'd plan for a CR test. This is one of the more important nuances of RPV in experimentation generally.
The decision rule
Declare a winner when RPV is significant at 95% AND the CR guardrail isn't breached. If RPV wins but CR drops below the guardrail, you've likely segmented your buyers — investigate whether the lift comes from a specific traffic cohort before rolling out.
Frequently asked questions
No. If the variant only changes copy, layout, or visual style without touching anything related to price, bundles, quantity, or upsells, CR is fine and reaches significance faster. Switch to RPV the moment the variant can plausibly move AOV.
Plan for 1.5-2x the duration of an equivalent CR test. Revenue per visitor has higher variance because a few large orders can swing the mean, so you need more data to reach the same confidence level. Two full purchase cycles is a good minimum.
AOV is revenue per order — it only counts buyers. RPV is revenue per visitor — it accounts for both conversion rate and order value in one number. Use RPV as the win condition; use AOV and CR as diagnostic metrics to explain why RPV moved.
No. Use order subtotal (pre-shipping, pre-tax). Shipping rules and tax rates vary by region, and including them adds noise that has nothing to do with what your landing page changed. Subtotal is the cleanest revenue signal.
Net refunds out of the variant's revenue total, attributed to the visitor's original assignment. If your test window is 14 days and refunds take 30 days to settle, hold the final readout for the full refund window or apply a historical refund-rate adjustment.
Most teams set a guardrail of 15-25%. A bundle variant lifting RPV 12% while dropping CR 8% is a clean win. A variant lifting RPV 4% while dropping CR 30% is a segmentation signal — you've probably found a cohort that loves the bundle while alienating everyone else.
Use first-order RPV as the primary metric and projected LTV as a secondary. Subscription LPs often have variants that change trial-to-paid economics, so you also want to track activation rate and second-order revenue over a 60-90 day window.
Meta and Google can shift audience composition mid-test, which moves RPV independently of your landing page. Monitor cohort consistency (CPM, CPC, audience overlap) during the test, and pause the readout if traffic mix changes materially between variants.
Below ~1,000 orders per variant, RPV readouts are unstable — a handful of large orders dominates the mean. For low-volume pages, either consolidate traffic to fewer variants, extend the test window, or test on CR first and validate the winner on RPV over a longer post-launch window.
Most legacy A/B tools (VWO, Optimizely) show CR by default and require custom revenue goals to surface RPV. Newer e-commerce-native platforms read order data directly from Shopify or WooCommerce and report RPV as a first-class metric, which avoids the manual revenue-event plumbing.
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.