Why E-Commerce Ads Are More Expensive and Less Effective -- And What the Best Operators Are Doing About It

You're not imagining it.

Your cost per acquisition is higher than it was two years ago. Your reported ROAS is lower.

Campaigns that performed reliably in 2022 require more budget to achieve the same results in 2025. The dashboards that used to show you clearly what was working now give you numbers that don't reconcile with your actual revenue.

This is not a creative problem, a targeting problem, or an account structure problem. The advertising environment in e-commerce changed structurally, across multiple platforms simultaneously. Most sellers are still using a 2021 mental model to interpret 2026 data.

Here is what actually changed, why it changed, and what the operators handling it best are doing differently. None of it is mysterious once you see the structure. It's just not being explained clearly.

What Is Blended ROAS and Why Does It Matter?

Blended ROAS is total revenue across all channels divided by total advertising spend across all channels for a given period.

If your business generated 150,000 pounds in revenue and spent 30,000 pounds on ads across all platforms, your blended ROAS is 5.0x. Regardless of what any individual platform reports it delivered.

This is the number that reflects what your business actually produced. It can't be inflated by a platform claiming credit for a conversion it didn't drive.

Four Forces That Changed E-Commerce Advertising Simultaneously

The deterioration in advertising performance that most e-commerce sellers experienced from 2022 onward wasn't driven by a single event. Four structural forces converged in the same window.

Apple's iOS 14.5 App Tracking Transparency framework rolled out in April 2021. It required apps to request explicit permission before tracking users across other apps and websites.

Tracking opt-in rates on iOS settled between 20% and 35% depending on app category. For social apps like Facebook and Instagram, opt-in rates ran toward the lower end. Meta lost the ability to directly observe the post-click behaviour of 65 to 80% of its iOS user base.

This affected everything downstream. The quality of the algorithm's training data for conversion optimisation.

The accuracy of Meta's reported ROAS. The precision of interest and behaviour targeting.

The effectiveness of lookalike audience modelling.

The full impact wasn't immediate. Meta's algorithm continued on accumulated historical data for roughly 12 to 18 months before signal degradation became visible in campaign performance.

By mid-2022 the effect was apparent to sophisticated advertisers. By 2023 it was impossible to ignore.

The important nuance: iOS 14.5 did not necessarily cause actual conversion rates to fall. What it caused was the observable, reportable conversion rate to fall.

For many brands, actual revenue from Meta-driven customers held relatively stable while reported ROAS declined. Decisions to cut Meta spend based on platform-reported ROAS were sometimes wrong.

Without a cross-channel view, there was no way to know that.

The advertiser base expanded permanently at the same time. The e-commerce surge of 2020 to 2021 brought a significant wave of new advertisers onto Meta and Google.

Many survived and continued advertising. Ad auctions are real-time second-price auctions.

More bidders competing for the same inventory produces higher clearing prices. CPM inflation from 2020 to 2024 was significant across most e-commerce categories.

The cause was straightforward: more advertisers competing for the same ad inventory.

This compounded the attribution problem. Sellers were getting less accurate data about which impressions worked at exactly the same time they were paying more for each one.

Amazon then systematically expanded sponsored placement inventory. In 2018, a typical Amazon search results page showed 1 to 2 sponsored product placements before organic results.

By 2025, the standard layout shows 4 to 6 sponsored placements above the fold on most competitive search pages. More sponsored products throughout organic results.

More at the bottom of the page.

Organic reach declined structurally. Amazon's Q1 2025 earnings report showed advertising revenue growing 19% year-over-year to 17.2 billion dollars while retail sales grew 7%. Those numbers are not coincidental.

TikTok created a new attention economy that redistributed where advertising investment had to go. TikTok's growth changed where consumer attention lives.

That changed where advertising investment needed to go. That increased competitive pressure in every existing channel as brands spread budgets wider.

Brands that built TikTok creative capability gained access to a significant and underpriced audience during TikTok's growth phase. Brands that didn't faced increasing competition in their existing channels from competitors bidding more efficiently across a wider platform set.

All four forces are structural. None of them reverses.

That's the uncomfortable truth most ad agencies aren't telling their clients. Partly because saying so would require admitting they don't have a simple fix.

The Attribution Problem Nobody Has Actually Solved

The practical consequence of iOS 14.5 is that no major ad platform now accurately reports what it's actually driving in your revenue.

Meta uses Modelled Conversions to estimate the conversions it can no longer directly observe. These are statistical estimates, not measurements.

Google uses its own attribution model, entirely contained within Google's ecosystem. It assigns no credit to Meta, email, or any touchpoint outside Google's visibility.

Amazon's reporting attributes conversions to Amazon traffic only. It ignores customers who discovered your product through a YouTube comparison video and came to Amazon specifically to buy.

When a customer sees a Meta ad, clicks a Google result, then searches Amazon and buys — Meta reports a conversion. Google reports a conversion.

Amazon may report one too. Three platforms claiming credit for one transaction.

This is the normal state of multi-channel advertising in 2025.

Add up the conversions reported by Meta, Google, Amazon, and TikTok separately and the total routinely exceeds your actual revenue. Every platform is claiming credit for the same customers.

Sellers who optimise based on individual platform ROAS are making budget allocation decisions on systematically inflated data. They cut spend on channels that report poor performance but are actually driving awareness that converts elsewhere. They scale channels that report high ROAS but are primarily capturing demand created by other channels.

Blended ROAS cannot be gamed by a platform claiming credit for a conversion it didn't drive. It reflects what the business actually produced relative to what it actually spent. It should be the primary number guiding your budget allocation decisions.

What Changed About Creative

The format shift in effective e-commerce creative is real. Brands whose early success was built on polished product photography are the ones feeling it most acutely.

Meta's feed is now majority video. Static image ads appear in a feed dominated by Reels, Stories, and video content.

Users conditioned to video scroll faster past static images. The attention threshold is lower for static content because users have learned to skim it.

The bar for static creative to earn attention is materially higher than in 2021.

TikTok users have developed sophisticated pattern recognition for content that looks produced or ad-like. The platform's algorithm rewards content that drives completion rate and shares. These metrics correlate with genuinely useful or entertaining content, not polished promotional material.

The most effective e-commerce creative on TikTok looks like organic content. A real person using a product in a real context.

A genuine problem demonstrated and solved. High production values are actively counterproductive on TikTok.

They signal "this is an ad." That's the kiss of death on that platform.

Google's Performance Max campaigns have changed the control model. Advertisers supply creative assets and audience signals. The algorithm decides where and how to show them across Google's full network: Search, Shopping, YouTube, Display, Discover, and Gmail.

Performance Max can produce efficient overall ROAS numbers while quietly absorbing branded search spend on terms you would have ranked for organically. Reported ROAS looks great. But those conversions would have occurred regardless.

Across all three platforms, the consistent finding from 2022 to 2025: creative testing volume is more predictive of performance than production quality. The number of distinct concepts tested per month matters more than how polished each one is.

The winner is almost always unexpected. The only way to find it is to test.

The brands winning on paid social typically produce 15 to 30 creative variations per month. Not 3 to 5.

What the Best Operators Actually Do

The e-commerce brands whose advertising is performing well in 2025 share operational practices that are materially different from brands struggling with rising costs and declining efficiency. The differences are not primarily about which platform they're on. They're about how they use data to make decisions.

They measure blended ROAS as their primary metric. When Meta's reported ROAS drops, they check whether blended ROAS has moved.

If blended is stable, they don't cut Meta spend. They investigate why the attribution gap has widened.

If blended has also declined, they have a real signal and act on it.

They triangulate platform data with first-party signals. Post-purchase surveys asking "how did you hear about us?" provide signal that no platform can manipulate.

Tracked through their own order data: new customer revenue as a percentage of total revenue. It tells them whether they're genuinely acquiring new customers or mainly selling to existing ones.

Monthly cohort retention rates indicate whether acquired customers have genuine lifetime value or are low-quality single-purchase buyers.

They treat creative as a volume game with rigorous performance tracking. A defined number of new concepts introduced per week.

A consistent framework for measuring performance, typically 3 to 5 days of spend above a minimum threshold before evaluating. Clear criteria for scaling winners and cutting losers.

A process for feeding creative learnings back into the next testing cycle. This is not complicated.

Most e-commerce brands don't do it systematically.

They have a cross-channel view that shows the whole business. When Amazon ad costs spike, they can see immediately which products are unprofitable and whether shifting spend elsewhere is viable. When Meta efficiency drops, they can see whether Google or TikTok has available capacity at their target ROAS.

That cross-channel view isn't achievable through a combination of individual platform dashboards. It requires a data layer that sits above all platforms, pulls data consistently, and presents a unified format that enables real decisions.

Alpomi connects Google Ads, Meta, Amazon Seller Central, TikTok, and Shopify into exactly that view. Updated hourly. Blended ROAS, cross-channel attribution, and margin by product calculated automatically.

When the advertising environment changes again, and it will, the operators with this view will make adjustments in hours. The others will find out weeks later, in their payout.

See how Alpomi's unified dashboard works and book a demo

Frequently Asked Questions

Why did Meta ad costs increase so much after 2021?

Meta CPM inflation from 2021 to 2025 was driven by two compounding forces. Apple's iOS 14.5 framework degraded Meta's targeting signal.

It removed the ability to track the majority of iOS users across apps without explicit consent. Each impression became less efficiently targeted, requiring more spend to reach the same quality of audience.

The expansion of the advertiser base from the e-commerce growth of 2020 to 2021 increased auction competition for the same ad inventory. More bidders, same inventory, higher clearing prices.

Why does my total platform-reported ROAS add up to more than my actual revenue?

This is attribution overlap: multi-touch double-counting. Meta reports a conversion.

Google reports a conversion. Amazon may report one too.

Three platforms. One transaction.

Three platforms claiming credit for one transaction. This is the normal state of multi-channel advertising in 2025.

It's why individual platform ROAS is misleading as a primary optimisation metric and why blended ROAS is more reliable for budget allocation.

What creative formats work best for e-commerce advertising in 2025?

On TikTok, content that looks organic, genuine product demonstrations, authentic reviews, problem-solution formats in natural settings, consistently outperforms polished ad creative. On Meta, short-form video with motion in the first 2 seconds outperforms static images in most categories.

On Google Performance Max, a broad range of creative assets across all formats gives the algorithm more to work with. Across all platforms, testing volume, the number of distinct concepts per month, is more predictive of performance than production quality.

How has Amazon's advertising changed for sellers?

Amazon expanded its sponsored placement inventory significantly from 2018 to 2025. From 1 to 2 sponsored results per search page, to 4 to 6 above the fold in most competitive categories.

This structurally reduced organic visibility, making paid advertising a required operational cost rather than an optional growth investment. Amazon's advertising revenue growing 19% year-over-year in Q1 2025 while retail sales grew 7% reflects this structural shift.

What is Performance Max and should I use it?

Google Performance Max allocates budget across the full network using machine learning based on your creative assets. Search, Shopping, YouTube, Display, Discover, Gmail.

It can be highly efficient at finding conversion opportunities. But the opacity makes it hard to know which placements drive real conversions versus which just capture branded demand that would have arrived regardless.

Most sophisticated e-commerce advertisers run PMax alongside brand-exclusion standard Shopping campaigns to separate branded and non-branded performance.

How do I calculate my true blended ROAS?

Add your total revenue across every channel, Amazon, Shopify, direct, wholesale, for a given month. Add your total advertising spend across every platform for the same month.

Divide revenue by spend. That number, calculated monthly and tracked consistently, is the only advertising efficiency metric that can't be gamed by platform attribution.