DIGITAL MARKETING

One foundation, one habit: find the records that answer a question, count them by written rules, and say plainly what they show.

1. We brought the history into one place. In a reporting warehouse, we joined the live store and its archive into one order timeline back to 2015, refreshed daily. Each customer is followed across both by the email address on their orders. A new question now starts from the history, not a fresh export. First we settled what counts as revenue: a subscription’s schedule isn’t money until it bills. Every load is checked against the business’s own totals.
2. We kept personal details inside the store. A retention policy needs old data deleted; we went further. The analysis only needs to know that the same customer is the same customer, so before anything leaves the store’s server, each email address becomes a code that can’t be turned back into it, and names, phone numbers and street addresses are dropped. A check on every export stops the load if an email slips through; it has caught two. Once a month, anyone with no order in three years and no active, paused or on-hold subscription has even that code destroyed, so nothing in the warehouse can tie their history back to them. We compared customer counts, lifetime revenue and every monthly cohort before and after. Nothing moved. This is the warehouse, where the reporting happens; the store’s own records are separate work.
3. We settled how a new customer is counted. The store’s own labels overcounted paid search and the campaign tags undercounted it, so we read the page each customer actually arrived on. Then we wrote the rules down, detailed enough that someone new can reproduce the numbers exactly: a new customer is the first time an email address ever orders, and a renewal is never anyone’s first order. Rebuilt in the warehouse, the method closely matched the hand-built report. A gap found later was corrected in the open. The signals it reads are the subject of our attribution case study.
4. We turned repeat asks into reports. Figures that meant a hand-written query or an email to a developer, like which subscriptions hold a given meal, now sit in the store’s admin for MegaFit’s team to run. When our resubscription analysis found customers who cancel and return more than once, MegaFit asked to keep watching them. We built that into their reports and checked it against the analysis.
5. We answered the question behind the question. In mid-2026 MegaFit asked whether subscriptions were dropping. Our first pull seemed to say yes, until we traced the drop to how cancellations were dated. We fixed that at the source and said so. They weren’t dropping. The concern wasn’t baseless; it was aimed at the wrong number. The total kept growing while the number actually being billed went flat. New sign-ups had slowed since January while cancellations held steady, so the pressure was on acquisition, not retention. And long-standing subscribers were going on hold and staying there, never counted as canceled and no longer billing. Whether to reactivate or close them is MegaFit’s call.