Case study
Ninety-two out of every hundred users never said yes
Motorhome accessories e-commerce · The audit that explained why the numbers didn't add up
A product priced at €7,41 was reaching analytics as 7.410.000.
It isn’t a typo. It is the data this e-commerce business had been sending to its tracking system for months, and the reason why no report ever matched the till.
This case study isn’t about rankings or campaigns. It’s about something far more boring and far more expensive: a store generating good revenue that can’t prove why. Let’s dive in:
92,2%
users without consent
×1.000.000
the price error in analytics
135
weekly dead clicks on the banner
+30%
of orders identified as recoverable
1. The brief
An online store selling motorhome and campervan accessories, with an extensive catalogue, operating across two markets —Spain and Portugal— and boasting an average order value (AOV) of €182. High volume: over three hundred thousand sessions per quarter, with the organic channel contributing nearly 72%.
The brief was not “bring us more traffic”. It was “explain to us why our numbers bear no resemblance to our reality.” Allow me to explain 🙂 when a store with this kind of volume fails to recognize its own figures, the problem is rarely strategy; it lies in the underlying plumbing. And to see it, you have to pop open the bonnet.
2. What we found
Why did 92% of their users not appear in any report?
Because they never interacted with the consent banner. Out of 35.200 new weekly users, only 2.742 —a mere 7,8%— interacted with it. The rest browsed, made purchases, and left without having said either yes or no.
And among those who did interact with it, 36% rejected tracking, compared to the industry standard of 10% to 15%. That figure had a technical explanation: the banner buttons were built as text links with a click event rather than actual buttons, resulting in 135 dead clicks per week —people clicking on something that did not respond.
And what about the rest of the tracking setup?
The technical audit uncovered a chain of compounding failures:
- Price multiplied by one million in the data layer, which completely broke the e-commerce schema.
- A duplicated pageview event: the tag was triggering from two separate container managers simultaneously.
- The data transmission function rewritten by hand in the site code, interfering with the tag manager.
- The add-to-cart event carried over products from the previous listing, thus logging what the user had viewed rather than what they had actually purchased.
- Two entire checkout funnel steps were missing, and the discount field was permanently sending a value of zero.
- 22.212 sessions with no source attributed in a single month —nearly 15% of total traffic.
It goes without saying that with a setup like that, no monthly report is worth the paper it’s printed on. Therefore, the first deliverable wasn’t a strategy: it was an exact inventory of what was broken, complete with sample code snippets right beside it.
3. What we delivered
Thirteen months of analytics and consultancy work, fully documented and dated:
- Full site analytics audit, including the consent banner diagnosis and the identification of duplicated tracking.
- Implementation review across the two coexisting analytics systems, detailing which event was failing and why.
- Data layer analysis, which led to uncovering the price bug and the corrupted cart data.
- JavaScript error report and banner impact report, combining heatmaps with actual user behaviour —which exposed the 135 weekly dead clicks.
- Campaign and shopping experience analysis in the Portuguese market, where two active campaigns had spent €594,27 with zero attributed conversions and were triggering search terms from a completely unrelated industry.
4. The result: knowing how much money is left on the table
An analytics project isn’t paid for in pageviews. It pays for itself through decisions that are no longer made blindly.
| Find | Figure | What does it involve? |
|---|---|---|
| Users without consent | 92,2% | Nine out of ten decisions were being made on a biased sample |
| Banner rejection rate | 36% | Compared to the usual 10–15%: a design issue, not an audience one |
| Weekly dead clicks | 135 | Users clicking a button that wasn't actually a button |
| Price error in the data layer | ×1.000.000 | The entire revenue report in analytics rendered useless |
| Unattributed sessions | 22.212 in one month | Nearly 15% of traffic could not be attributed to any channel |
| Ad spend with no measurable ROI | €594,27 | Two campaigns with zero attributed conversions |
| Quantified opportunity | ~80 orders/month | 30% more orders, at an average order value (AOV) of €182 |
The opportunity of 80 monthly orders stems from the friction analysis of the checkout process in the lower-converting market compared to the main market. This is an internal estimate based on observed data, not an achieved result, and was presented to the client as such.
I recommend carrying out a very simple test on your own store: take the number of orders reported by your e-commerce platform this month and place it next to the figure shown in your analytics. If the difference exceeds 5%, every decision you make based on those numbers is being made with the wrong information.
In summary: are your analytics telling you fairy tales?
Essentially, that is the question. And the answer usually lies in the plumbing, not the strategy.
Three double-checks for this week: Look at what percentage of your visitors interacts with your consent banner: if it is below 20%, your reports are describing a minority. Verify that the price reaching your analytics is the actual price, complete with its decimals and currency. Check how many sessions are logged as “unassigned” traffic: if it exceeds 10%, you can no longer tell where your business is coming from.
Remember, at the end of the day, it’s all about making sure your numbers actually match your till 😉
