Platform Forward GroupThe reporting stack nobody owns
Every estate has a reporting layer that grew instead of being built. How to inventory it, what it costs, and the rebuild pattern that actually holds.
Ask who owns the CRM and you get a name. Ask who owns reporting and you get a silence with politics in it. The reporting layer is where every estate keeps its unowned complexity: the Excel that pulls from the view that reads the replica that a stored procedure feeds at 5 a.m., and the one person who understood the chain left in 2021.
We treat reporting as its own inventory pass now, separate from applications, because the application inventory never finds it. Reports hide in mailboxes, scheduled tasks, BI folders named "FINAL v3", and department shares. The pass is a week of work:
- Harvest the schedules. Every emailed report, every BI subscription, every cron that writes an xlsx anywhere. The count will be three to five times what anyone guessed.
- For each: who receives it, and did anyone open it this quarter. Distribution logs answer the second question and the answer retires a third of the list on the spot. Nobody mourns a report. They mourn the argument about retiring it, which the open-rate data ends before it starts.
- For the survivors, trace the chain to source tables and write the owner column. Blank cells are the finding, same as always.
What the survivors usually reveal: the same numbers computed three ways by three departments, each slightly differently, each defended. That is not a tooling problem, and buying a new BI platform will faithfully migrate all three versions of the truth. Fix the definitions first, then the plumbing.
On the rebuild, steal a working order of operations rather than inventing one. The regulated reporting rebuild is the pattern we point at: kill the manual assembly first, standardize the definitions as the automation forces the question, and let the tooling decision come last, after the governed Tableau platform has an owned pipeline to sit on. Read the order: automation, definitions, then tools. Every failed reporting modernization we walk ran that list backwards, bought the tool first, and gave the old chaos a nicer front end.
Price it honestly and the business case writes itself: count hours per month of humans assembling numbers, multiply by loaded cost, and put it beside the week of inventory above. We have never seen the inventory lose.
The 5 a.m. stored procedure is not your problem. The silence when you ask who owns it is your problem. Take the week.