E-commerce runs on tracking — what shoppers browse, abandon, and buy feeds the ads, the recommendations, and the retargeting that drive revenue. It’s also the most heavily enforced corner of the privacy regime, because the cookie banner is where data collection meets the public, and regulators have spent years fining the same predictable failures. The uncomfortable part for most online stores: the banner that looks fine, converts well, and was copied from a competitor is very often the one a regulator would reject on sight.
Tracking Starts Before the Sale
An online store collects personal data long before anyone reaches checkout. Analytics scripts, advertising pixels, session recorders, and retargeting tags all fire as the page loads, building a profile of a visitor who hasn’t bought anything and may never. Non-essential trackers like these need prior consent; only the genuinely essential ones — the shopping cart, fraud prevention, load balancing — are exempt. The line between “essential” and “nice for marketing” is narrower than most stores assume, and almost everything in the typical ad and analytics stack sits on the wrong side of it.
Consent and the Ad Stack
For stores running Google’s advertising or measurement products on European traffic, the banner is only the front of a pipeline — the choice a shopper makes has to actually reach the tags. Google’s Consent Mode is the mechanism that carries it; a banner that records a clean choice but never transmits it leaves the store in the worst of both worlds: non-compliant on paper and losing the measurement data it broke the rules to collect. The collection problem and the conversion problem turn out to be the same problem.
In December 2020 France’s regulator fined Amazon €35 million over the cookie banner on its French store. The problem was the oldest one in the book: advertising cookies were placed on shoppers’ devices the moment they arrived — including when they landed by clicking an Amazon ad on another site — before any consent, behind a banner that simply announced “by using this website, you accept our use of cookies.” That implied-consent model, the regulator held, met none of the requirements: no prior consent, no real choice, and no clear information on what the cookies did or how to refuse them. The case wasn’t a relic, either — in 2025 the same regulator fined the fast-fashion retailer Shein €150 million for cookie failures, a reminder that this is live enforcement, not settled history. For e-commerce, the banner is the single most-tested compliance surface there is, and “everyone’s banner looks like this” has never been a defence.
Looks Compliant vs. Is Compliant
Pattern-matching from real store reviews — the gap between a banner that looks fine and one that holds up tends to follow the same shape:
| Looks compliant | Is actually compliant |
|---|---|
| ✗ “By using this site, you accept cookies” | ✓ A real choice before non-essential tags load |
| ✗ Tags fire the moment the page loads | ✓ Tags held until the shopper opts in |
| ✗ “Accept all” with no easy way to refuse | ✓ Reject as prominent as accept, on the same screen |
| ✗ One toggle for ads, analytics, everything | ✓ Granular consent per purpose |
| ✗ Banner consent that never reaches the tags | ✓ Consent propagated to the ad stack via Consent Mode |
| ✗ Marketing emails to everyone who bought | ✓ Consent or a valid soft opt-in, with easy unsubscribe |
| ✗ “Everyone’s banner looks like this” | ✓ A banner built to the rules, not to the neighbours |
Compliance Doesn’t Have to Kill Conversion
The objection is always the same: a genuine reject button will tank the numbers. In practice the trade-off is overstated — a clear, honest banner with a real choice tends to perform close to a manipulative one, and the shoppers who decline tracking were rarely the high-intent buyers anyway. What is not overstated is the fine. The store that bets its banner against the regulator is risking a number with seven or eight digits to protect a conversion delta that usually turns out to be small.
Final Thought
E-commerce lives and dies on tracking, and tracking is exactly what regulators watch most closely. The banner isn’t a cosmetic detail to copy from a competitor — it’s the most-enforced surface in the entire regime, and the failures that get fined are the same ones, over and over: consent assumed, choice denied, reject hidden, purpose obscured. Getting it right is neither hard nor expensive. Getting it wrong has a price list, and it runs into the hundreds of millions.
The test: load the store fresh and answer three things — do any non-essential tags fire before a choice is made, can a shopper refuse as easily as accept on the same screen, and does the banner actually say what’s collected and by whom. If any answer is no, the banner is a finding waiting for a complaint.
Frequently Asked Questions
Yes. Prior consent means the analytics, advertising, and personalisation tags wait until the shopper opts in. Only genuinely essential cookies — the shopping cart, fraud prevention, load balancing — are exempt, and the line is narrower than most stores assume.
Yes. Symmetry is one of the most-fined failures. If accepting is one click and refusing takes two screens, the consent is not freely given and regulators have repeatedly held that against retailers.
Consent Mode is the mechanism that carries the shopper’s choice to the ad and measurement stack, so it is necessary if you run those tools. It is not a substitute for a compliant banner — the choice still has to be valid before it is propagated.
In practice the impact is overstated. A clear banner with a real choice tends to perform close to a manipulative one, and the shoppers who decline tracking were rarely the highest-intent buyers. The conversion delta is usually small; the fines are not.