Email Variant Testing: 22 Abandoned Cart Combinations, a 15x Revenue Spread, and the Data Behind Both

Every abandoned cart benchmark you have ever read reports one number. Klaviyo publishes a single blended revenue-per-recipient figure of $3.65 for cart flows. Omnisend reports a single aggregate click rate around 8%. Salesforce, Barilliance, everyone else, same pattern. One metric, one use case. But email variant testing at the variant level tells a radically different story, because abandoned cart is not one use case. It is 22.

Zembula’s Q4 2025 Benchmark, drawn from 6.2 billion email opens across brands on the platform, breaks abandoned cart into 22 distinct signal combinations. Each combination pairs the cart abandonment event with zero, one, or two additional behavioral or contextual signals (loyalty status, price drops, low stock, countdowns, coupons, BNPL eligibility, browse history, and more). The performance spread across those 22 combinations is a 15x gap in Revenue Per Mille: from $31.42 to $469.65, normalized to the same $100 average order value. The reason no competitor publishes this data is that they don’t measure at variant level. They can’t. And that gap in measurement creates a gap in every optimization decision that follows.

The Performance Spread: $31 RPM to $470 RPM Inside One Use Case Family

Here is what the data actually looks like when you stop blending. All figures are RPM normalized to $100 AOV from Zembula’s Q4 2025 Benchmark.

Cart Alone, the variant most brands are running, generates $135 RPM. That’s the baseline. Cart + Loyalty + Price Drop, a three-signal combination, generates $469.65 RPM. That’s a 3.5x improvement over the baseline, and it’s 15x the worst-performing variant (Cart + BNPL at $31.42). Every dollar of the gap between those bars is money that exists inside your abandoned cart program right now, uncaptured, because the signals aren’t being stacked.

As McKinsey reported in 2021, companies that excel at personalization generate 40% more revenue than average players. The gap between Cart Alone and Cart + Loyalty + Price Drop is 248%. That’s not a marginal personalization improvement. That’s a different league.

Why Email Variant Testing Shows the Best CTC and Best RPM Are Different Combinations

This is the finding that should change how you think about cart recovery optimization. The highest-RPM variant in the Zembula benchmark is Cart + Loyalty + Price Drop at $469.65 RPM. The highest click-to-conversion variant is Cart + Coupon + Countdown at 27.9% CTC. Those are completely different signal combinations.

A brand that optimizes only for CTC (click-to-conversion rate) will miss the revenue leader by 56%. A brand that optimizes only for RPM will miss the conversion quality leader. Both metrics are correct. They just measure different things, and the signal combinations that maximize each one are not the same. This is why Revenue Per Mille is the metric that wins CFO meetings, and why CTC remains the metric that tells you whether your content is relevant. You need both, measured at the variant level, to make real portfolio decisions.

The baseline CTC for an entire email is roughly 2.5%. Cart Alone runs at 19.6% CTC, already strong. But Cart + Coupon + Countdown pushes to 27.9%. The signal combination tells you who is responding and why. Without email variant testing at this granularity, you’re flying blind on which audiences are actually converting.

How Signal Stacking Compounds Toward Higher-Intent Audiences

Each signal you layer on top of the cart abandonment event does something specific: it filters the audience down to a higher-intent subset. Cart Alone addresses everyone who abandoned. Cart + Loyalty filters to abandoners who also have loyalty status (and therefore repeat-purchase behavior). Cart + Loyalty + Price Drop further narrows to loyalty members whose abandoned item has dropped in price, a triple signal that screams purchase readiness.

The audience for a three-signal variant is smaller than the audience for Cart Alone. But the RPM is 3.5x higher. That math works because email block analytics at the variant level let you see this tradeoff clearly. You’re not choosing between reach and relevance as a binary. You’re running a portfolio where different variants serve different audience segments, and each one earns its slot based on measured RPM.

The compounding effect is real. Two-signal variants in the Zembula benchmark (Cart + Coupon, Cart + Price Drop, Cart + Loyalty) cluster between $218 and $267 RPM. Three-signal variants (Cart + Coupon + Low Stock, Cart + Loyalty + Price Drop) jump to $301 and $470 RPM. The third signal doesn’t add 10%. It adds 20% to 75% on top of already strong two-signal combinations.

What BNPL and Low Stock Teach About Signal-Audience Fit

Conventional email wisdom says “add urgency to abandoned cart emails.” The benchmark data says that advice is dangerously incomplete.

Cart + BNPL, the combination that layers buy-now-pay-later eligibility onto the cart signal, produces the worst RPM in the entire abandoned cart family at $31.42. Why? BNPL signals price sensitivity and often low immediate purchase intent. The subscriber needs financing, which means the price point is already a friction point. Adding urgency to that audience doesn’t overcome the objection. It amplifies it.

Meanwhile, Cart + Countdown (pure urgency) runs at $279 RPM, and Cart + Coupon + Countdown (urgency layered on a discount) produces the highest CTC at 27.9%. Same urgency mechanic. Wildly different outcomes. The lesson: urgency is a multiplier. What it multiplies matters more than the urgency itself. Layer it on a high-intent signal (coupon redeemers, loyalty members) and it accelerates. Layer it on a low-intent signal (BNPL, cold browse) and it pushes back.

Cart + Low Stock is interesting in this same way. At $235 RPM, it outperforms Cart Alone by 74%. Scarcity works when there’s already intent. But combine it with a coupon (Cart + Coupon + Low Stock at $301 RPM), and you get the urgency-discount combo that performs even better. The Nielsen Norman Group’s eye-tracking research on banner blindness explains part of this: subscribers learn to ignore static, repetitive content. Rotating variant combinations, dynamically selected by signal match, are the counter to that.

The Portfolio Question: Which Variants Are Running and What’s the Cost of the Gap

According to the Baymard Institute’s 2025 data, the global average cart abandonment rate is 70.22%, with mobile running between 79% and 85%. That represents hundreds of millions of high-intent abandoners per year across retail. Most of them are receiving a single-variant cart recovery email.

This is the portfolio question that email variant testing forces you to answer: which of the 22 possible combinations are you running today? And what is the compounding cost, per 1,000 impressions, of the variants you’re not running?

If you’re running Cart Alone at $135 RPM and you could be running Cart + Loyalty + Price Drop at $470 RPM for the segment that qualifies, that’s $335 per thousand impressions left on the table for that audience. Over a year, across your full cart abandonment volume, the gap compounds fast. The mature question isn’t “is abandoned cart working?” It’s “how much of the available performance am I actually capturing, and which signal combinations would change the math?”

This is why turning abandoned carts into personalization opportunities is a portfolio management discipline, not a one-and-done setup. Every time your data layer gains a new signal (loyalty tier, price drop feed, inventory status), a new variant becomes possible, and the RPM spread widens.

Email Variant Testing Requires Variant-Level Attribution Infrastructure

None of this analysis is possible if your measurement stops at the block level or the flow level. Variant-level attribution means measuring RPM and CTC per signal combination, not blended across all cart emails. It means knowing that Cart + Coupon + Countdown converts at 27.9% CTC while Cart + BNPL converts at a fraction of that, and being able to act on the difference.

Zembula’s architecture makes this work through a few mechanisms. Open-time content selection picks the right variant per subscriber at the moment of open, not at send time. A single URL supports all 22 combinations without ESP template changes or HTML weight penalties. And 7-day click-based revenue attribution tracks each conversion back to the specific variant that drove it. You can read more about how to measure email ROI performance with click-to-conversion as the core metric.

Without this infrastructure, you’re stuck with what every other benchmark report publishes: a single blended number for abandoned cart. That number is correct as far as it goes. It just hides a 15x performance spread underneath it, and it makes portfolio-level optimization impossible. As Zembula has written before, return on spend is core to long-term success, and you can’t optimize return on spend if you can’t see which variants are earning it.

Key takeaways

• Abandoned cart is not one use case. It is a family of 22 signal combinations, each with distinct RPM and CTC performance. Treating it as one number masks a 15x revenue spread.
• The best-converting variant and the highest-revenue variant are different combinations. Cart + Coupon + Countdown leads on CTC (27.9%). Cart + Loyalty + Price Drop leads on RPM ($470). Optimizing for only one metric means missing the other leader.
• Signal stacking compounds. Three-signal variants generate 20-75% more RPM than two-signal variants. Each additional signal filters toward a higher-intent audience.
• Urgency is a multiplier, not a strategy. Cart + BNPL ($31 RPM) and Cart + Coupon + Countdown (27.9% CTC) use the same urgency mechanic on different audiences with 15x different outcomes.
• Email variant testing at the combination level requires variant-level attribution. If you measure at the block or flow level, the 15x spread stays invisible. Variant-level RPM and CTC, measured per signal combination, are the prerequisite for portfolio optimization.
• Every missing variant has a measurable cost. If Cart Alone runs at $135 RPM and Cart + Loyalty + Price Drop runs at $470 RPM, the gap is $335 per 1,000 impressions for every qualified subscriber not receiving the higher-performing combination.