1,000 stores and 16 distribution centres is the scope Eroski put around a single forecasting decision. The Spanish supermarket co-operative named RELEX Solutions as its forecasting and replenishment provider, with Deloitte supporting implementation. As of November 9, 2026 the scope is public only through trade round-ups, and no measured result has been published. What the announcement does not include is a single measured result, and that gap is the most useful part for anyone planning a similar move.
What a 1,000-store replan has to prove
Replacing a legacy replenishment system is not a software swap. Every store order becomes an output of a model rather than a habit, and store managers who used to override weekly now override hourly unless the proposal quality is good enough to be trusted. That trust is earned store by store, not systemwide at cutover.
The second thing a replan has to prove is that store and distribution centre plans are looking at the same demand. A promotion can empty one location while cases sit at another, and the only way to catch it is to plan store orders and DC capacity together rather than in sequence. Eroski says coordination between the two is the point of the project.
The third is that the model survives fresh. Short-life categories punish a two-day forecast error much harder than ambient ones, and the replenishment parameters that work for packaged goods will over-order chilled lines until waste forces a manual reset. Most programmes discover this in the first chilled aisle they pilot, not in the specification.
Three comparable cases, only two with numbers
The vendor's published cases give a sense of what a realistic outcome looks like. Two of them report operating results; the third, Eroski, has announced a rollout. Reading a target as a result is the most common mistake in this category of project.
| Case | What was measured | Reported outcome | Status |
|---|---|---|---|
| Medovia | Product availability, June 2024 to May 2025 versus the following year | Up 1.3 percentage points; about 91 percent of proposals left unchanged | Operating |
| Angulas Aguinaga | Share of forecast volume used without planner edits | About 80 percent, roughly 4,000 SKUs | Operating |
| Eroski | Store and DC replenishment across about 1,000 stores | No measured improvement published | Rollout |
The Medovia figure is the more instructive of the two. A 1.3 percentage point availability gain sounds small until it is multiplied by the number of order lines, and the 91 percent unchanged proposals matter just as much, because they say where planner time went: the unusual orders, not the routine ones.
Where replanning usually breaks
Store labour is the constraint nobody models
A replenishment proposal that arrives after the shelf has been faced is a proposal that gets ignored. Stores run on a fixed labour budget, and the receiving window, the facing routine and the delivery slot have to line up before a better order quantity can be executed. Projects that skip this step produce accurate plans that stores cannot act on.
The countermeasure is unglamorous: publish the proposal reliability by store for the first eight weeks, and let store managers mark the lines they do not trust. That list is the fastest route to the parameter that is wrong, because it is written by the people who face the shelf. Review it weekly and close the loop by telling stores what changed.
Fresh lines need their own parameters
Waste and availability also pull in opposite directions during the transition. Tightening order quantities improves freshness and drops availability; loosening them does the reverse. Reporting both on one page, weekly, is the only way to see which way a parameter change actually moved the business.
Master data deserves its own budget line. Item, location, lead time and shelf capacity records decide whether a model can produce a usable proposal at all, and they are usually dirtier than the programme assumes. A retailer that cleans them first spends less on configuration later, and stops the first quarter being spent on arguments about which record is right.
For suppliers the practical consequence arrives with the first ordering cycle after go-live. When a model sets the order, a listing without reliable case pack, lead time and shelf life data gets ordered conservatively, and conservative ordering looks exactly like a distribution gap from the brand's side.
FAQ
Is the Eroski project a measured result?
No. It is an announced rollout with a defined scope of about 1,000 stores and 16 distribution centres. No post-implementation improvement has been published yet.
Why does 91 percent of proposals left unchanged matter?
It shows where planner time goes. If routine orders are accepted as generated, the team can concentrate on promotions, new lines and unusual stores.
What breaks a replenishment replan most often?
Store labour and master data. A plan stores cannot execute, or a model fed with wrong lead times, fails regardless of forecast quality.
How should waste and availability be tracked together?
On the same weekly page. Tightening orders improves freshness and lowers availability, so reading either number alone will point the wrong way.
