# Case Study: The Customer Database That Was One Setting Away From Being Usable | Crank Source: https://wearecrank.com/case-studies/skincare-consent-retention-creative-2026 A fully built customer database sitting unused in the advertising platform, tracking that only ever saw people who said yes, and a clean test that isolated the real problem. --- ![Skincare case study: The Customer Database That Was One Setting Away From Being Usable](/images/case-studies/skincare-creative.png?dpl=dpl_2uJjggmfov9ibhArygujfRBb3pax) Skincare, 2026 Audit # The Customer Database That Was One Setting Away From Being Usable A fully built customer database sitting unused in the advertising platform, tracking that only ever saw people who said yes, and a clean test that isolated the real problem. Challenge The customer database sat fully built in the email platform and was never switched on for advertising, one unaccepted sync setting kept the business's own data from reaching targeting or suppression at all, while tracking itself only ever saw visitors who accepted cookies. Performance was being judged only on the people who accepted tracking, so the true figure was unknown in both directions, and no campaign could be judged on its own economics, one ran 23 creative variants against roughly 50 events needed for any single one to stabilise. Approach Accept the audience-sync terms so the customer database held in the email platform can reach advertising for both targeting and suppression, and switch on the platforms' own consent modelling so visitors who decline tracking are represented rather than simply absent. Then separate existing customers from new ones across the account, and deliberately refocus effort toward acquisition once the two can be told apart. Outcome Revenue grew significantly and return on spend fell, because more of the budget was deliberately going toward winning new customers, who cost more per sale than selling again to someone who already buys, the arithmetic of the choice, not a decline in performance. The brand grew overall volume as a result, and is positioned for greater customer value over time. 01 ## Measurement that only saw the people who said yes Measurement was partial and consent-dependent, so a material share of activity was invisible and unmeasured, performance was being judged against only those who accepted tracking, not everyone who actually bought. Server-side tracking was implemented and functioning correctly, but only captured data for visitors who had already consented; anyone who declined was never sent through anonymously, because the platform's own recovery option, letting Google or Meta model the gap left by declining visitors, had simply never been switched on. Not a bug. A normal setup that had never been extended. 02 ## The customer data that was never plugged in The email platform held the first-party customer data, but the audience-sync terms with the advertising platform were never accepted, so none of it could flow through to targeting or suppression, the plumbing was built and switched off at a single setting. It looks, from the outside, like a much bigger data problem than it actually is. 03 ## Spend without separation Acquisition, retention and lifecycle activity all ran through the same campaigns with no exclusion or suppression logic, so no campaign could be judged on its own merits. Both major shopping campaigns had the incremental value of a new customer versus an existing one set to $1, with no data basis given, too small a number to meaningfully shift the bidding algorithm toward acquisition, so the setting existed but did nothing. No abandoned-basket audiences or campaigns existed on either platform, and no replenishment activity had been built despite the products having natural repurchase cycles. 04 ## A clean test that isolated the real problem One campaign, cold prospecting only, existing customers and subscribers already excluded, delivered $208.40 cost per acquisition against a 0.32x return. Because the exclusions were already correctly in place, the loss couldn't be blamed on audience contamination from existing customers, which pointed the diagnosis straight at the offer and creative instead. Separately, a single campaign ran 23 creative variants simultaneously against a platform requirement of roughly 50 optimisation events to stabilise, spend spread too thin for any variant to reach that threshold, and a version of the same pattern found elsewhere in this evidence: more variants isn't better if none of them get enough data to mean anything. 05 ## What switching on the missing settings changed With the audience sync accepted and consent modelling switched on, the business's own customer data could finally reach targeting and suppression, and tracking stopped being blind to everyone who had declined it. Once existing customers were separated from new ones, effort was deliberately refocused toward acquisition, the reason the blended return on spend fell even as revenue grew. That is the arithmetic of a choice, not a decline. 06 ## What this means for you Once you can tell new customers from existing ones, a falling blended return becomes a decision you have made rather than a problem you have. You cannot make that decision at all while the two are mixed together, and a business that cannot separate them will read its own growth as failure. ## What a full-stack consent and creative audit found, and changed $208.40/0.32x Cost per acquisition and return, isolated cold-prospecting test 23 Creative variants in one campaign, vs. \~50 events needed to stabilise 1 setting Distance between the customer database and being usable in advertising On Audience sync and consent modelling, both switched on, closing the gap Drawn from Crank's account audit conducted in 2026 for this skincare 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)