Researchers physically audited nearly 370,000 inventory records across 37 stores at one major retailer and found 65% didn't match what was actually on the shelf. Not a rounding error, not a handful of outliers, roughly two-thirds of the records the automated replenishment system was making decisions from. Their modeled estimate of the resulting damage: more than 1% of North American sales and more than 3% of gross profit, lost to a forecasting system that was working exactly as designed on top of input it had no way to know was wrong.
That's a specific, fundable number. "Garbage in, garbage out" is true and nearly useless as a budget argument, because it describes a principle, not a decision, and a strategy meeting doesn't fund principles. It funds "this forecast is running on 65% bad input and it's costing us real margin." A later field experiment tested the fix directly: RFID-driven automatic record correction across a set of treatment stores cut inventory-record inaccuracy by about 26%, and a related study found stockout reductions of 21% to 36% depending on category. The number and its fix are both real, even though they come from one retailer and don't net out every implementation cost.
Move from hygiene to decision risk
The strongest case for quality work starts with the decision, not the dataset. What is this data about to drive, what would make the result wrong in a way that actually matters, and what does it cost to find that out later instead of now? That answer is what should determine which checks get priority. A minor formatting defect and a broken customer key are not the same risk, and treating them as though they compete for the same afternoon of engineering time is how the broken key waits eighteen months to get fixed.
That framing also makes the prevention case visible in a way "cleaning data" never does. The retailer's forecasting system wasn't broken. It was doing exactly what it was built to do, faithfully, on input nobody had checked. The work isn't abstract hygiene. It's reducing the odds that a forecast, a valuation, a launch decision, or an automated action gets built on top of something whose actual behavior nobody in the room can explain.
The budget question
Instead of asking whether the organization can afford a quality program, ask what decision it's already making without one, today, this quarter, with real money behind it. That question gives the work an owner, a deadline, and a consequence, the three things a generic hygiene initiative never has and a line item in someone's forecast always does.