Stop paying for work nobody uses, rebuilds or double-checks. That sentence sounds obvious. The hard part is finding it, because most of that cost never appears on a data invoice. It sits in business teams' calendars, in licences that renew quietly, and in products that were built for a decision nobody makes anymore.
Why the invoice is the wrong place to start
Ask a CFO what the company spends on data and the answer is usually the cloud bill, the tool licences and the team payroll. Those numbers are real, and they are the smaller part. The larger part is the time other people spend working around what the data team delivers: rebuilding a report in a spreadsheet, checking a number before using it, asking an analyst for a cut the dashboard should have given them.
In one engagement the visible data budget was a fraction of the real figure. Once the business hours were counted at payroll rates, the company was paying roughly €930,000 a year for data work, and about €611,000 of it could be stopped. It took three weeks to measure. Nobody had hidden anything. Nobody had counted it.
Look in three places
1. Manual reporting hours that a trusted number would remove. Find the reports people rebuild every week or month by hand. Count the people, the hours and their loaded hourly cost. This is what I call FTE Debt: the payroll you spend on workarounds. It is usually the biggest line and the least visible one.
2. Licences and pipelines nobody opens. Pull usage for every dashboard, tool seat and scheduled job. Anything barely used by the people it was built for, for a full quarter, is a candidate. This part you can check against supplier invoices, so it is the easiest to defend in front of a CFO.
3. Products with no business decision attached. For each major data product, ask one question: which decision changes if this disappears? If nobody can name one, the product is a cost without a return. Include what it costs to keep alive: incidents, support requests and the small changes that keep landing on it.
Check each item against something auditable
A list of things to stop is only useful if the CFO can trust it. Tie every line to a source: payroll rates for hours, supplier invoices for licences, ticket counts for support. Where a number is an estimate, say so and label how confident you are. A first rough number with an honest label beats a precise one that arrives in six months.
Stopping is a decision, not a clean-up
Most companies agree in principle and change nothing, because stopping something has an owner and a cost. Make it explicit. Name who approves each stop, set a date, and tell the people who use the product what replaces it. Then add one rule going forward: nothing new enters the portfolio without naming what it replaces. Without that rule the list grows back within a year.
Where to start this week
Take the three most expensive recurring reports in your company. For each, ask who builds it, who rechecks it and who actually decides something from it. If you cannot answer all three, you have found your first stop.
If you want to put a first number on it, the FTE Debt calculator takes about five minutes. If you want a structured version with your own team, that is exactly what the "What should we stop paying for in data?" workshop does.