Automation repeats a process. If the process counts cancelled appointments as attended visits, it will repeat that mistake just as reliably as a correct calculation. Start with a report you can check.
Begin with a bounded export
Choose a completed reporting period and a manageable set of questions. Appointments, new patients and care plans can be a useful starting point when those are the questions the owner needs answered. You do not need to solve every future reporting requirement in the first build.
Record the export date, filters and source reports used for comparison. Keep the data requirements tied to the management question; avoid collecting extra patient detail simply because it is available.
Reconcile before refining the visuals
Compare the totals with the source system for the same period and filters. When they differ, look for status rules, duplicate records, date boundaries, missing history or changed records. Keep a short explanation for each resolved discrepancy.
Write down the rules that matter. Does “new patient” mean first booked or first attended? Does the monthly appointment figure use booking date or appointment date? A definition settled here should apply everywhere the measure appears.
Get the owner to use it
Ask the owner to answer a real question with the report. Can they identify what changed, investigate the underlying records and explain the result? That feedback is more useful than adding another chart because there is space for it.
Then choose the refresh method
A controlled file-drop process, scheduled export or API may be appropriate depending on source capabilities and delivery needs. Compare the manual effort, reliability, access requirements and what happens when the source changes. An API is a means of moving data, not proof that the reporting is ready.
Make failure visible
A repeatable process needs a last-successful-refresh time, checks for unexpected data and someone responsible for responding. Agree how missing files, changed columns and failed checks will be handled. A report that looks current while showing old data is difficult to trust.
The sequence is straightforward: validate the source, agree the measures, refine the report, then automate the repeatable work. Wedge’s automation strategy follows that order.