Almost every warehouse has a column in the item master that says “class A”, “class B” or “class C”. Someone filled it in one day, with good judgment, and there it stayed. The problem is that the column is a photograph, and a warehouse is a movie. The product that was a star in March may have been trickling out for months, and the one nobody cared about may now ship every single day. Meanwhile, each one is still stored wherever the old label said.
This week I want to explain how we solved that at Silodisa AI, because it is one of the simplest pieces to explain and one of the ones that saves the most meters.
01The class is calculated, not declared
In our warehouse system no product comes with its class preset. The ABCD class is calculated from what actually left the warehouse in the last 80 days. Not a forecast, not the sales team’s opinion, not a two-year-old label: real outbound shipments, lot by lot.
The rule is the usual one, Pareto, but with fresh data. Products are ranked from most to least movement and accumulated: the ones that together explain most of the outbound volume are A; the next block is B; the long tail is C; and the ones that did not ship even once in 80 days are D. That last class is the one that surprises people most. A warehouse always has more sleeping stock than it thinks, and much of it sits in the best locations.
The 80-day window is not set in stone. It is a parameter, like every parameter the agents use: it has a source, an administrator approves it, and any change is logged with who, when and what the previous value was.
02Why it matters in meters, seconds and pesos
In the first installment I explained that we turn every pallet move into meters, seconds and pesos, and that a pallet that takes three minutes to move costs MX$4.75 with a forklift at MX$95 an hour. The ABCD class is what decides how many times you will pay that MX$4.75 for the same product.
A hypothetical but realistic case: a product that ships twenty times a month, stored at the back of the farthest aisle on the fourth level, forces twenty long trips and twenty slow lifts. Near the staging area and on the first level, it is twenty short trips. The difference per trip looks small, a minute or two; multiplied across hundreds of products it becomes hours of forklift time every week that nobody sees, because it is spread out in little pieces.
03What the Put-away agent does with it
The Put-away agent (Acomodador) uses the calculated class to propose where each incoming pallet goes: aisle and level. For A items it proposes close and low; for D items, far and high. It does not decide; it proposes, with its name and its human owner’s name next to it, and the supervisor sees the suggestion and the reason.
For what is already stored in the wrong place, the Planner (Planificador) steps in. You ask it the way you would ask a person — “which A products do I have at the back?” — and it builds a re-slotting plan that states three things:
The rule that always applies
Nothing moves on its own. Someone with authority reviews, approves and releases the plan, and before a task reaches the forklift driver the system checks again that the pallet is still where the plan saw it.
04What recalculating taught us
Three things we learned as soon as we stopped trusting the label:
First: the class moves. Some seasonal products are A for three months and C the rest of the year. With the 80-day window the system moves them up and down on its own, and the Put-away agent proposes different things in March than in September. With a fixed label, that never happened.
Second: there are more D items than anyone wants to admit. Seeing them in a list, with their location and what that location is worth, changes the conversation: it is no longer “we should clean up the racks someday”, it is “these pallets occupy these locations that cost this much”.
Third: the number has to be open to challenge. Next to every suggestion there is a “I disagree” button. If the supervisor knows that a product that looks like a C today is going to take off next week because of a campaign, they say so right there, and the Put-away agent’s owner has to respond.
05The same calculation, outside the warehouse
We are taking this approach beyond the four walls: fleet, cold chain on the road, deliveries with a service-level clock, and replenishment at the point of consumption. The question there is identical: what is really being consumed, where and at what pace? Not what the catalog says or last year’s order. What actually shipped in the last 80 days.
Several of our numbers are still red, and pallet location is one of the slowest to turn green, because moving what is already stored costs money. But for the first time we know which pallets are in the wrong place, how much they cost us there every day, and how many days it takes for moving them to pay off. That is already a different warehouse.