Email vs Paid Ads: The 2.87x Problem and the Written 10x Guarantee Behind Smart Banners

Two numbers frame the email vs paid ads conversation going into 2026. Average ecommerce ROAS fell to 2.87 in 2025, and the median sits at 2.04, which means half of all ecommerce brands are buying revenue at less than a 2:1 return. Against that backdrop, my company puts a 10x ROAS floor in writing for email Smart Banners. This post is about why one of those channels can support a written guarantee and the other cannot, and what that asymmetry should tell you about where your next marketing dollar goes.

I am the CEO of Zembula, so apply whatever discount you think is fair. The argument does not rest on my numbers, though. It rests on a principle any CFO will recognize: a guarantee is an underwriting decision, and nobody underwrites risk they cannot measure. The ad industry has been selling performance for two decades and has never produced a written ROAS floor. That absence is data.

To be clear, this is not a case for killing paid media. It is a case about the marginal dollar, and about why Smart Banners give that dollar a measurable home.

Half of e-commerce now runs below 2:1

The 2.87 figure comes from Upcounting’s October 2025 analysis, and it is not an outlier reading. Eightx landed on the same 2.87 number across a dataset of more than 35,000 ecommerce brands. The median is worse: Triple Whale’s benchmark data puts the median ecommerce ROAS at 2.04, so a full half of the industry operates below 2:1. Even paid search, the highest-intent channel in the mix, averaged just 2.26x in WebFX’s 2025 analysis.

The driver is cost inflation, not creative fatigue. Upcounting notes that Meta CPMs hit $10.88 in Q1 2025, up 19.2% year over year, and puts it bluntly: “If your ROAS targets haven’t adjusted for rising costs, you’re measuring success with outdated math.” Returns compressed across 13 of the 14 industries Upcounting tracked. This is arithmetic every CEO already sees in the quarterly paid media review.

Credit where it’s due: what paid media does that email cannot

Paid media manufactures net-new demand. Email cannot introduce your brand to a stranger, and no retention math changes that. Stop feeding the top of the funnel and your list stops growing, which erodes the owned-audience argument over time. Paid earns a large budget for a real reason.

The ad platforms also built the measurement culture the rest of marketing now aspires to: daily ROAS and CPA reporting with spend accountability at the campaign level, and always-on testing baked in. That discipline is the standard email should be held to, which is the argument we made in Email Is a Performance Marketing Channel, and the Math Proves It. The budget context makes this urgent. Per Gartner’s 2025 CMO Spend Survey, marketing budgets are flat at roughly 7.7% of company revenue while paid media remains the focal point of spend. Flat budgets plus rising media costs means the marginal-dollar question is no longer optional.

ROAS is a revenue ratio, not a profit metric

Upcounting again: “ROAS only tells you one thing: how much revenue came back for every dollar you spent on ads. That’s it. It doesn’t tell you if you’re actually making money.” Their worked example lands the point. A 4:1 campaign on $10,000 of spend nets roughly $8,800 in actual profit once product costs, processing fees, and overhead come out, and a “worse” 3:1 campaign with a better cost structure can out-earn it.

Now apply that logic across channels. In paid, media cost is consumed with every impression, so the biggest expense in the channel scales in lockstep with revenue. Email’s cost base is mostly fixed. You pay for the ESP and the tooling whether you send or not, so incremental revenue falls almost entirely to contribution margin. We worked this out line by line in Email Unit Economics: How Smart Banners Make a 3x ROAS Worth 5x More Than Paid Ads. At identical reported ROAS, the email return is worth several times more in contribution than the paid return. ROAS parity is not economic parity, and every email vs paid ads comparison should start there.

Why Smart Banners can carry a written 10x guarantee (and ads cannot)

A written performance guarantee is a measurement artifact. Before anyone underwrites a floor, they need a return number they trust enough to bet margin on. Paid media cannot produce one. Since Apple’s App Tracking Transparency rollout, ad platforms can see an estimated 40 to 60% of the conversions they drive (Ruler Analytics). Attribution models disagree with each other by design: platform-reported, last-click, and data-driven numbers routinely tell three different stories about the same campaign. Platform lift studies exist, but they are run by the party selling you the media, on an identity graph with ATT-sized holes in it, and you cannot independently rerun them. Nobody signs a floor on top of numbers like that.

Email supports a fundamentally cleaner experiment because the brand owns the identity layer. Randomize your own list, collapse the personalization pixel for the holdout group, keep every other element of every send identical, and compare UTM-scoped email channel revenue between the two groups. Same list, same sends, a true randomized controlled test that ad platforms cannot offer on their own inventory. We detailed the mechanics in The Collapsed-Pixel Holdout. Zembula’s longitudinal version of this test, assigned on open across every send carrying Smart Banners, typically reaches 95%+ statistical significance in about four weeks and can be rerun any time a CFO asks.

A finance-grade number makes the risk quantifiable, and quantifiable risk is insurable. That is the entire reason the Smart Banners guarantee can exist in writing, with three outcomes that all favor the customer. If the measured return comes in under 10x, the price drops until it lands at 10x. At 10x, the promise is simply kept. Above 10x, the customer keeps all of the upside. As of 2025, Zembula customers average 15x, with results as high as 41x, and the two Smart Banner slots in broadcast sends produce a holdout-validated 15 to 20% lift in email channel revenue. Smart Banners themselves are personalized image blocks that decide their content per person at the moment of open.

The reallocation math: one point of ad budget vs the Smart Banners line

CRM budgets typically run 5 to 10 times smaller than paid budgets, and that asymmetry is the whole opportunity. Take a brand spending $20 million a year on paid media and $2 million on the email program. Moving one percentage point of ad spend, $200,000, costs the ad team a rounding error on a channel averaging 2.87x. The same $200,000 is a 10% budget expansion for email, enough to put Smart Banners and Smart Kickers into 100% of broadcast sends, pointed at an owned audience with first-party identity and a return that can be holdout-verified. One ledger barely notices the move. The other steps up a level.

Rising acquisition costs sharpen the trade. Shopify’s Global Commerce Report tracked average merchant CAC climbing from $274 to $318 year over year. We walked through the reallocation trigger in The $318 Problem and the planning-cycle math in Retention Marketing ROI. And note what the ad industry itself already believes: Meta Custom Audiences, Google Customer Match, and every lookalike model in your stack are seeded from email lists. The channels where your costs keep rising run on the first-party asset you already own.

Run the comparison on your own P&L

You do not have to take any vendor’s word for this, mine included. The comparison runs in about twelve weeks.

4 Week Test: Run the email channel-level holdout. Around four weeks gets most senders to 95%+ significance on email channel transactions. At the end, you have a randomized, holdout-proven return on the Smart Banners investment to put next to paid’s modeled ROAS in your next planning meeting. Only one of those numbers will survive a CFO’s questions.

Key takeaways

• Average ecommerce ROAS fell to 2.87 in 2025 and the median is 2.04 (Upcounting, Triple Whale). Half of ecommerce buys revenue below 2:1.
• ROAS is a revenue ratio, not a profit metric. On contribution margin, an email return at the same reported ROAS is worth several times more than the paid equivalent.
• A written guarantee is a measurement artifact. Paid attribution (40 to 60% conversion visibility post-ATT, models that disagree by design) cannot support underwriting a floor. Email’s randomized channel holdout can.
• The Smart Banners guarantee has three outcomes and all favor the customer: price drops until the return lands at 10x, the promise is kept at 10x, and the customer keeps everything above it. Zembula customers average 15x as of 2025, with results up to 41x.
• One point of ad budget is a rounding error to the paid ledger and a step-function upgrade to email.
• Twelve weeks of work produces a holdout-proven email number. Paid media cannot produce its equivalent.