"We increased budget on a high-CVR ad and revenue did not grow." I hear this from EC operators every month. CVR (conversion rate) is improving but the monthly revenue is flat — or worse, declining. The team thinks the campaign is working because CVR is up, but the P&L disagrees.
The cause is almost always the same: CVR is being used as the sole judgment metric, and AOV (average order value) is dropping in the background. Revenue Per Session (RPS) — the composite of AOV × CVR — is the metric that actually moves with revenue, and it often moves opposite to CVR. If you only watch CVR, you systematically miss the cases where the campaign is killing your top line.
Note on terminology: I use Revenue Per Session (RPS) below. RPS is not a standardized industry metric in the way ROAS or LTV are — it is RevenueScope's core metric. I spell out the full term on first mention.
TL;DR
- RPS is an absolute revenue-efficiency metric; CVR is an intermediate purchase-rate metric. From RPS = AOV × CVR, when AOV drops, RPS can fall even as CVR rises.
- 3 patterns produce CVR-RPS divergence: bundle discounts, low-price funneling, and aggressive coupons. All three look like "CVR success" while burning revenue.
- A 2-axis quadrant chart of RPS and CVR turns ad budget allocation into a clean decision: prioritize Q2 (high RPS × high CVR), apply AOV-lifting to Q4 (the CVR trap zone), withdraw from Q3, scale Q1.
The CVR-RPS divergence problem
CVR and RPS measure different things, even though they sound related.
- CVR = Purchase Sessions ÷ Total Sessions (a percentage). It's an intermediate metric — the rate of conversion.
- RPS = Revenue ÷ Sessions (a currency value). It's an absolute metric — actual revenue per visit.
The two are linked by RPS = AOV × CVR, which means CVR is just one input. Concrete numbers: AOV ¥6,000 with CVR 2.0% gives RPS ¥120. If CVR climbs to 3.0% but AOV falls to ¥3,500, RPS becomes ¥105 — actually worse. The dashboard says "CVR up 50%" while revenue per visit dropped 12.5%. The CVR improvement does not imply a revenue improvement.
The reason this misjudgment is so common: GA4 surfaces CVR prominently in standard reports, while RPS is a derived metric that requires a small extra step to compute. Operators read what's in front of them.
The 3 traps where CVR rises while RPS falls
When CVR moves opposite to AOV (or sessions shrink in parallel), three specific patterns produce a fall in RPS.
Trap 1: Bundle discounts pull down AOV. A "10% off second item" promotion can push CVR from 2.0% to 3.5%, but AOV often drops from ¥6,000 to ¥3,500 in the process. The math: ¥3,500 × 0.035 = ¥122.5 RPS, versus the ¥120 baseline. A 42% AOV drop in exchange for a 2.1% RPS gain is not a successful initiative.
Trap 2: Low-price product funneling. A "¥980 trial product" funnel pushes CVR up sharply (2.0% → 6.0%, a 3x lift) but destroys AOV (¥6,000 → ¥1,200, a 5x drop). The result: RPS goes from ¥120 to ¥72, a 40% decline. Outside of a deliberate LTV-recovery model, the monthly numbers are bad.
Trap 3: Site-wide coupon distribution. 10-20% off coupons across the catalog look like a CVR win, but AOV drops by the discount amount and the math repeats from Trap 1.
The shared structure across all three: directional divergence between CVR and RPS. Treating "CVR up = success" guarantees you'll fall into at least one of these patterns.
The RPS × CVR quadrant: a decision framework
To make this visible at a glance, plot ad channels (or LPs, or product pages) on a 2-axis chart with CVR on the x-axis and RPS on the y-axis.
The four quadrants give you a direct action mapping:
- Q1: High RPS × Low CVR — high-AOV product hit, scale candidate. Increase budget.
- Q2: High RPS × High CVR — best revenue efficiency, top priority. Maximize allocation.
- Q3: Low RPS × Low CVR — largest improvement room, withdrawal candidate. Reduce budget or rebuild LP.
- Q4: Low RPS × High CVR — the CVR trap zone. AOV-lifting initiatives required.
The decision flow is three steps:
- Compute monthly RPS and CVR for each ad channel.
- Plot into the four quadrants.
- For Q4 channels, judge as "not eligible for budget allocation unless AOV is lifted" instead of "keep because CVR is high."
Q4 is the most easily misjudged. "Good channel because CVR is high" is the wrong read — most Q4 channels are sitting on Trap 1, 2, or 3 and need an AOV intervention before they deserve more budget.
How to apply this in monthly budget reviews
In monthly ad budget reviews, the order I recommend is:
- Revenue — outcome metric, overall picture.
- RPS — primary axis for cross-channel comparison.
- AOV and CVR — decomposition of why RPS moved.
- Sessions — inflow scale check.
Treat CVR as a supporting metric and never let CVR alone drive budget decisions. The migration from CVR-only judgment to RPS-as-primary is mostly a KPI configuration change in your dashboard — not a tooling overhaul.
The Japanese B2C EC market reached ¥24.8 trillion in 2024, and the quality of ad budget allocation directly drives growth. The teams that get this right at SMB scale (¥10-50M monthly revenue) tend to be the ones that graduate to enterprise scale, while the teams stuck on CVR-only judgment plateau there.
If you want the full version with formulas, references, and the GA4 implementation details, I wrote a longer article on it: RPS vs CVR: A Two-Axis Framework for EC Ad Budget Allocation.
What's the most surprising CVR-RPS divergence you've seen in your own ad data? Curious where Trap 1, 2, or 3 has bitten others.
