Ecommerce ad performance has been declining for two years straight, and tweaking your campaigns isn’t going to fix it.
According to WARC’s Future of Measurement 2026, the entire measurement system that ecommerce advertising runs on has broken down. The signals are wrong, the attribution is inflated, and most brands are optimising against numbers that no longer reflect reality.
Here’s what you need to know to survive this obstacle.
The Numbers Are Ugly And You Deserve To See Them
Average ecommerce ROAS in 2025 was 2.87:1, down 4% year over year. The median is 2.04:1. Half of all ecommerce brands are getting less than $2 back for every $1 spent on ads.
It gets worse as mid-market and large brands saw ROAS fall 9% last year. Google Ads ROAS dropped 10% across industries. Conversion rates fell 9.28%. CPCs rose 10–25% across nearly every vertical.
This is structural, driven by three things happening at once: platforms getting more expensive, tracking infrastructure collapsing, and measurement systems that were already behind the curve.
Your Ad Data Is A Lie And Privacy Killed It
Between 15 and 30% of your conversions are invisible to ad platforms right now. Not because of anything you did, but because the tracking system digital advertising was built on no longer works.
Safari blocked third-party cookies in 2020. Firefox did the same. Apple’s App Tracking Transparency killed cross-app tracking for 75–85% of iPhone users. Google spent five years promising a replacement, then abandoned the whole Privacy Sandbox project in October 2025. There is no replacement.
What this looks like in practice: someone sees your Instagram ad on an iPhone, clicks, browses on Safari, and buys two weeks later. Meta never sees that conversion. The cookie expired. The algorithm decides that audience doesn’t convert, and moves your budget away from them. You end up defunding your best customers based on incomplete data.
You can’t fix this with a pixel update. This is the permanent state of digital advertising now.
Google And Meta Are Both Claiming Credit For The Same Sale
Even if tracking were perfect, platform-reported ROAS would still be inflated.
Google’s Performance Max reports 3.5:1 to 5:1 ROAS. Incrementality testing — where you actually ask “would these sales have happened without the ad?” — consistently shows lower real returns. PMax takes credit for brand searches and retargeting conversions that were going to happen anyway.
Meta does the same thing. Their attribution windows overlap with Google’s. Both platforms claim the same sale. Your blended dashboard ROAS is higher than your actual return because two platforms are splitting one conversion and each calling it a win.
Meanwhile, costs keep climbing. Meta CPMs averaged $10.88 in Q1 2025 — up 19.2% year over year. Q4 runs 30–50% higher. The margin behind your “good” ROAS number is shrinking whether the dashboard shows it or not.
Smart Bidding adds one more layer: when every advertiser lets Google’s AI run their bids, the algorithms homogenise. Every player competes for the same signals in the same auctions. The inefficiencies that sharp buyers used to exploit are gone. Everyone gets average results, by design.
WARC’s Wake-Up Call: Stop Measuring The Wrong Things
WARC’s report identifies three shifts that matter for ecommerce specifically.
Measure outcomes, not proxies. Reach, clicks, and attributed ROAS are no longer reliable indicators of business results. The gap between what platforms report and what actually happened in your revenue has become too wide to ignore. Digital platforms are already moving toward real-time outcome-based optimisation. Most brands are still running on last-click attribution from 2018.
Be careful with AI measurement tools. AI is moving from backend data processing into active campaign decision-making. That’s not inherently bad, but if the input data is degraded — and it is — AI systems will optimise confidently toward the wrong outcomes. A black box running on broken data doesn’t produce better results. It produces faster, harder-to-detect mistakes.
Creative is now a measurable performance driver. WARC’s Future of Measurement 2026 argues that creative quality has been systematically undervalued because it was hard to measure. AI tools are changing that — making it possible to test assets at scale, predict performance before launch, and connect creative quality directly to conversion data. Brands that treated creative as a commodity and competed purely on targeting and bids are now at a disadvantage.
Four Things To Fix Before You Spend Another Dollar On Ads
Set up server-side tracking. Stop relying on browser pixels. Send conversion data directly from your server to ad platform APIs. Meta’s own data shows this recovers 8–19% more attributed conversions. That signal feeds the algorithm and improves optimisation. It’s the most immediate fix available.
Stop treating platform ROAS as your source of truth. Run incrementality tests. Hold back ads for a control group, compare results, and find out what revenue your advertising is actually generating — not what Google and Meta are claiming credit for.
Use marketing mix modelling. It’s no longer just for enterprise budgets. MMM gives you a view of advertising effectiveness that doesn’t depend on platform tracking or cookie-based attribution. It’s making a comeback because it’s one of the only measurement approaches that still works in the current environment.
Invest in creative properly. Most brands refresh creative after performance drops. That’s already too late. Build a systematic process for testing at volume — different hooks, formats, angles — and rotate before fatigue sets in.
The Platforms Work. Your Faith In The Dashboard Shouldn’t.
The channels aren’t broken. Global ad spend is heading toward $1.3 trillion in 2026. The money is still there, and the platforms still work.
What’s broken is the measurement layer underneath, the tracking, the attribution, the dashboards. Brands optimising against those numbers are making budget decisions based on fiction.
