# Case Study: Ten Things Wrong, None Visible From Inside the Advertising Platform | Crank Source: https://wearecrank.com/case-studies/gaming-full-audit-2026 Unverified de-duplication, basket recovery that existed only on paper, and email open rates a third of the retail norm, a full-stack audit of one account. --- ![Gaming case study: Ten Things Wrong, None Visible From Inside the Advertising Platform](/images/case-studies/gaming.png?dpl=dpl_2uJjggmfov9ibhArygujfRBb3pax) Gaming, 2026 Audit # Ten Things Wrong, None Visible From Inside the Advertising Platform Unverified de-duplication, basket recovery that existed only on paper, and email open rates a third of the retail norm, a full-stack audit of one account. Challenge Purchase events were being sent from both browser and server without confirmed de-duplication, basket recovery existed 'on paper' with nobody having verified it actually ran, and email open rates sat at a third of the retail norm, each one invisible until someone went looking specifically for it. Reported sales could have been counted twice, or understated where visitors declined tracking, so every budget decision rested on a number no one had checked, and validated purchases weren't reaching the advertising platforms either, so their own optimisation was running on partial information too. Approach Implement correct tracking and de-duplicate the purchase events arriving from both browser and server, so a sale is counted once. Feed the corrected conversions back to the advertising platforms so every sale is attributed rather than only those the platform happens to see. Then split audiences into new and existing customers, so the true cost of winning a customer can be separated from the cost of selling to someone who is already buying. Outcome With conversions counted correctly and returned in full, recorded cost per acquisition fell and return on spend rose, the account had been understating its own performance, not overstating it. Separating new from existing customers gave the business, for the first time, a real acquisition cost to plan against. 01 ## A measurement system nobody had verified Purchase events were being sent from both the browser and the server without confirmed de-duplication, meaning reported sales could be counted twice before any budget increase. It was also unclear whether anonymous conversions were being recorded for visitors who declined tracking, meaning reported performance may have understated reality by an unknown margin in the other direction. Validated purchases, high-value customers and database records were not being returned to the advertising platforms either, so their own optimisation was running on partial information, improving attribution quality here also improves how well the platforms can find genuinely valuable customers, not just report on them. 02 ## Activity that existed only on paper It was not confirmed whether abandoned-basket and dynamic retargeting flows existed or functioned at all, despite this being among the highest-returning activity available in the account. Email open rates of 10 to 12 per cent were recorded against a typical retail range of 30 to 40 per cent, pointing to a segmentation and database quality problem rather than a content one. Very few products carried visible reviews, affecting both buyer confidence and click-through rate in shopping results. 03 ## Structure that was never built No distinct campaign structures existed for acquisition, retention or cross-sell across either advertising platform, budget could not be directed at growth as opposed to people who were returning anyway. Product identifier completeness and category mapping were unverified across the feed, directly limiting shopping and catalogue quality. Campaigns ran a single headline, text and description combination, giving the platform nothing to optimise between. And lead volume was being optimised for without any review of downstream quality, conversion rate or eventual value. 04 ## What correcting the count actually revealed The advertising platforms gained complete information to optimise against, rather than a partial view, once corrected conversions were fed back in full. And splitting new customers from existing ones gave the business a basis for deciding what to spend on winning customers versus keeping them, a decision a single blended figure had never been able to support. 05 ## What this means for you Fixing measurement does not always reveal worse numbers. Here it revealed better ones, the account had been judging itself on an incomplete count. And until you separate new customers from existing ones, you do not know what winning a customer actually costs, so you cannot sensibly split budget between acquisition and retention. ## Ten findings, one account, and what correcting them changed 10-12% Email open rate found, vs. 30-40% sector norm 1 Headline/description/text combination running, per campaign, before the audit 0 Sales double-counted, once browser and server events were de-duplicated Newvsexisting Customers separated for the first time, giving a real acquisition cost to plan against Drawn from Crank's account audit conducted in 2026 for this gaming retailer. Client not named. ## Want to know what this looks like in your own account? These findings came from asking straightforward questions before any budget moved. We're happy to do the same for yours. [ Talk to Wearecrank](https://wa.me/447457416922?text=Help%20my%20business)[Get in touch](/contact) [← Back to all case studies](/case-studies)