Governance business case: turning bad data into money

Whoever asks for a data governance budget usually presents the problem in technical vocabulary: duplication, inconsistency, lack of standards. Whoever approves the budget decides in a different vocabulary: loss, risk, working capital, fines.

The translation between the two is the business case. Without it, the request competes at a disadvantage against projects that already come with a number.

Where bad data costs money

Five sources cover most cases:

• Fragmented purchasing. The same item recorded three times prevents consolidating volume, and losing negotiating power is a direct loss of discount.
• Duplicate inventory. A duplicated item generates duplicated safety stock, and tied-up capital has a known monthly cost.
• Rework. Hours spent checking, correcting, and manually reconciling have a measurable hourly value and volume.
• Rejected invoices and returned deliveries. A wrong address and wrong tax data generate freight, re-invoicing, and customer service costs.
• Regulatory exposure. Here you don't estimate savings, you estimate risk: probability times impact.

How to measure without a six-month project

Don't try to measure everything. Pick one source, measure it well, and extrapolate honestly, stating your assumptions.

An example of a defensible slice: take the hundred items with the highest spend, identify how many have a duplicate, and calculate the price difference between purchases of the same item under different codes in the last year. That gives you an observed loss, not an estimated one, and observed loss is what sustains the conversation.

Then state the extrapolation out loud: "we measured 3% of the base and found R$ X; if the rate holds, the total is on the order of Y." The person approving respects the caveat; what destroys credibility is a round number with no origin.

What goes on the single page

The current situation with one measured number. What happens if nothing is done, over twelve months. What is proposed, in three sentences. The cost, split between people, tool, and time. The expected gain with the assumption made explicit. And the metric that will prove whether it worked.

One page. The appendix can have thirty.

Promise what you can sustain

An approved business case becomes a commitment. Promising an 80% reduction in duplication in three months guarantees approval and defeat in the next quarter.

Prefer a modest target with a short deadline and a stated review. A governance program is sustained by an accumulation of credible deliveries, not by a big number at the start.

Who presents

If possible, whoever feels the pain: procurement, tax, operations. The request made by the area that loses money carries a different weight than the same request made by the data team. The governance team prepares the numbers; the process owner defends them.