DSI reads one thing. The operation does another on holiday week.
DSI, days sales of inventory, is average inventory divided by daily sales at cost. The formula is honest. The standalone reading is not.
The number answers a finance question: how many days of sales are sitting in stock right now. Thirty days of DSI means capital turns once a month. For the CFO, it is one of the cleanest lines on the dashboard.
What it measures well
DSI captures trapped capital. Up from 30 to 45 days, there is extra cash sleeping in the distribution center. Down to 20, capital is lean.
For that question, the indicator works. It compares well month over month, and well chain against chain of the same format.
What it hides
DSI is an average over a period. And averages erase peaks. Retail shelf-out does not live in the average; it lives on the eve of the holiday, the first week of the month, the promo Sunday.
A chain can run DSI of 28 days, stable for a year, and 9% shelf-out every holiday eve. The inventory number says “healthy” because it looks at the 90-day average. The shelf on Friday says something else.
The low DSI the CFO celebrates is sometimes the exact cause of peak shelf-out. Stock that is too lean has no slack for the day sales triple.
The pair that is missing
DSI alone lies by omission. The pair that makes it honest is availability at the shelf, OSA, measured on peak days, not on the average.
Read together, the two tell the whole story. Low DSI with high OSA at peak is real efficiency. Low DSI with OSA that collapses on the holiday is risk dressed as discipline.
The implementation note
A forecasting project that targets DSI without watching the pair delivers lean capital and shelf-out on the worst day. The criterion has to separate base stock from peak stock, and measure each against its own day.
It is the same aggregate reading that misleads in MAPE and in stock cover. Solving it is criterion design, not another dashboard.
Take the inventory metric that has sat on the CEO’s slide for six months and barely moved. Is it stable because the operation is stable, or because the measurement is too coarse to show what moves?
Send me its name and how it is measured today. In one hour I will send back a one-page audit: the measurement quality problem, the pair that should be tracked alongside, and the size of the work if the audit confirms a gap. If it confirms, the two-week Diagnóstico scopes the rebuild.