The buyer trusted the spreadsheet over the system. He was partly right.

We arrived at a regional grocery chain on a Tuesday morning. A few dozen stores, tight working capital, a commercial director with a problem already named.

The problem, in his words: the dry-grocery buyer ignored the replenishment system and ordered from his own spreadsheet. Leadership wanted us to make the buyer “trust the system”. They asked for training. We asked for three days to look at the data first.

What we found that nobody had said

The system was not ignored out of stubbornness. It was ignored because it was learning wrong.

The historical sales feeding the model did not separate sales from shelf-out. When an item hit zero on the shelf, the system read zero sales and concluded: low demand. Next cycle, it ordered less. Less became more shelf-out, which became less recorded sales. The model was teaching itself to run out.

The buyer saw this without naming it. His spreadsheet corrected by hand what the system sank. The rebel leadership wanted to tame was the load-bearing wall holding the operation up.

What we changed

No new model at first. First, flag shelf-out in the history, so the system stops treating an empty shelf as lack of demand.

Then, capture what was in the buyer’s head. He knew which suppliers ran late, which city market emptied the store, which promotion pulled the neighboring shelf. We turned that into a model input, instead of treating it as noise.

The spreadsheet was not thrown away. It became a source of knowledge, not the system’s enemy.

The number that moved

In eight weeks, dry-grocery shelf-out fell from about 7% to near 3%. The buyer moved to reviewing exceptions, not typing orders line by line.

The honest part: the number also fell because a problem supplier was replaced in the same period, and because seasonality helped. It was not all model. It never is.

What we got wrong

We lost the first two weeks to arrogance. We walked in calling the work “replacing the buyer’s spreadsheet”. He heard it, shut down, and stopped cooperating. Rightly.

We had to go back and reframe it: the goal was to make the system as smart as his spreadsheet, with his knowledge inside it. From there he opened up. We lost two weeks learning not to treat the operator as the obstacle. It is the same mistake we describe in replacing the spreadsheet with automated forecasting.

What we left behind

A simple weekly ritual: the buyer reviews the exceptions the system flags, and the system learns from his corrections. A definition of shelf-out by day of week on the twenty fastest movers, instead of aggregate MAPE. The internal team runs it today, without us.

Think about your own “rebel buyer”. The person who ignores the official system and solves it their own way. Before correcting the behavior, it is worth asking what they are seeing that the system is not.

If this shape of gap sounds familiar, between the number on the dashboard and what the floor team tells you, describe in one paragraph what is happening. If the shape confirms a useful observation, we step inside the operation for two weeks and leave with a one-page document your team keeps: what we saw, what the metric missed, and the three changes that close the gap. That is how we work.