The starting point was a single spreadsheet that several people edited at once. It held roughly a decade of patient and revenue history and it was, genuinely, the most valuable asset in the business — which is precisely why it was dangerous. One bad paste would have destroyed years of records, and nobody could ask it a question more complex than a filter.

We resisted the temptation to design the ideal system first. The first phase was simply getting the data somewhere safe and structured: modelling patients, visits, production and referral sources properly, importing the history, and reconciling the imported totals against the spreadsheet until they matched exactly. Until the client trusted that the numbers had survived the move, nothing else was worth building.

With a clean foundation, the analytical work became possible. Lifetime value was computed from actual production and visit history rather than estimated. Churn was defined against dental recall behaviour — a patient overdue relative to their own interval — instead of borrowed from a subscription model where it does not apply. Retention was cut by cohort so the business could see whether it was genuinely improving rather than reacting to a monthly number.

The finding that changed the engagement was the churn analysis surfacing roughly $800K of annual revenue attached to patients drifting away unnoticed. That number reframed the project internally: it stopped being an IT tidy-up and became a commercial priority, which is usually what determines whether a platform like this gets adopted or quietly ignored.

Multi-user access with proper roles came next, which sounds mundane and mattered enormously — it removed the single point of failure and let the team act on the analysis simultaneously rather than queueing behind whoever had the file open.