Cellar Signals watches for inactive inventory, closing house service windows, concentration and by-the-glass exposure, then orders what it finds by what is actually on the signal. A caution worth forty thousand sits above an alert worth six hundred. Ten minutes a month, and a clean month is reported as a clean month rather than padded.
Sort the cellar by cost tied up in lots with no depletion in twelve months. Take the top twenty. For each one ask whether it is a listing problem, a pricing problem or a buying mistake. Listing problems go to the glass or to a by-the-glass Coravin position. Pricing problems get repriced against what the market actually pays now. Buying mistakes get sold or poured to staff, and you stop buying that wine.
That exercise takes an afternoon and finds most of the money. Aivin runs it monthly and shows the working, which is a different proposition from selling you the idea.
A finding with no money attached is not raised. A clean month says so and stops. Nothing you dismissed comes back without Aivin saying it is coming back and why.
Every threshold is a setting with the money currently behind it shown next to it, because a signal a house cannot tune is a signal it learns to ignore by the second month.
The GM wants beverage cost under thirty and I am at thirty-four.
Indirectly, and the honest mechanism matters. Aivin does not renegotiate with distributors. It finds the capital sitting in lots that have not moved in a year, the bottles past their window still priced as though they were young, and the by-the-glass exposure nobody has costed. Repricing and pouring those is where the points come from.