Three warehouses, one hub, no reorder point

The consolidation was decided. The planning logic it depended on did not exist.

An 8 to 10 week sprint. One senior analyst and two juniors. A €1 billion global coffee and food company, in its traded goods segment.

Three warehouses were being consolidated into a single regional hub. That decision was made and it was a good one. What did not exist was the thing the decision quietly depended on, which is a defensible answer to how much of each product line to hold, and when to reorder it.

Three conditions made the gap real rather than theoretical.

Fragmented warehousing. Three separate sites carried overlapping stock keeping units with inconsistent stock control and limited visibility between them. Nobody could see the same product across all three at once.

No link between the inventory data sets. The central warehouse and the regional sites did not share a view, which meant there was no basis on which to set an optimal reorder point or a safety stock level for anything.

Short shelf life. This is what makes the problem harder than a textbook safety stock exercise. Holding cost is normally capital tied up. Here, holding too much of the wrong line does not tie money up, it destroys it, because the product expires. Excess stock and stockouts were both occurring, which is the signature of a system with no model underneath it.

So the honest description of the starting position is not that the company was doing inventory planning badly. It is that it was doing it without an instrument, in a situation that had just become materially harder.

They could have built it. You build one of these once in a career.

Our own record of the engagement states the reason in one line, and it is worth quoting because it is more honest than most: the operations teams lacked the time and the analytical tools to build a data-driven inventory model internally.

Take the two halves separately, because they are different problems and only one of them is about capability.

The people who know these product lines best are the people running them. That is not a coincidence. It is the reason the work never starts.

The tools. A shelf-life-constrained reorder model across hundreds of product lines and several sites is not a spreadsheet you knock up between other duties. It is a piece of analytical machinery. A planning team builds one of these once, perhaps twice, in a career, and the first one is always the expensive one to learn on.

The weeks. Even with the method in hand, someone has to open every product line and establish its real demand pattern, its real lead time and its real shelf life. That is not clever work. It is patient work, and it is precisely the work that a team running a live supply chain cannot absorb without dropping something else.

None of that says the client was incapable. It says the two scarce things were uninterrupted weeks and a method someone had built before. That is the whole trade, and it is worth being plain about it rather than dressing it up.

Three figures, and two words that decide what they mean

Designed is not achieved. The 97% is a service level the model is built to hold. It is a target with arithmetic behind it, not a measurement of what the operation subsequently delivered. Nothing in this engagement observed a post-implementation service level, and if we ever quote it as one, challenge us.

Potential is not banked. The 10% is disposal loss the shelf-life modelling identified as removable. Whether it comes out depends on whether the reorder guidelines are followed when a season turns and somebody wants to hold more.

We labour this because the alternative is how a modelled number quietly becomes a claimed result, and because this is the page on our own site where the qualifiers are cleanest. Keeping them clean is worth more than the figures are.

The senior made four judgement calls. The juniors opened two hundred product lines.

It is a fair question why a company with a competent supply chain function buys this rather than running it internally, so here is the actual division of labour rather than a description of it.

What the senior decided. What service level to design to, and to write it down as a design target rather than let it drift into being reported as an outcome. Whether shelf life enters the model as a hard constraint or as a cost, which changes every answer downstream. Which product lines genuinely need a model and which are fine on a simple rule, because modelling everything is how these projects die. And what the simulation should be allowed to vary, which is the difference between a useful stress test and a random number generator.

What the juniors did. Opened more than 200 product lines one at a time and established the real demand pattern, lead time and shelf life for each. Sat in more than five workshops. Worked through more than 10 scenarios. This is weeks of unglamorous work and it is the reason the model is worth anything, because a beautiful method on bad parameters is just a confident wrong answer.

What stayed behind. Two tools, the reorder guidelines at product line level, and a team trained to run them. Not a deck. The 70% reduction in analysis time is the measurable trace of that, and it is the number that keeps paying after everyone has gone home.

Four things, named because a case study that names none is an advertisement

Whether the designed service level is achieved in operation. The 97% is what the model holds. Confirming it would need a post-implementation measurement, which was not part of this work and which we have not seen.

Whether the waste reduction is realised. The 10% is potential identified by modelling. Realising it depends on the guidelines being followed in the months when somebody has a good reason to override them.

The exact number of product lines. Our own internal record states it two ways on consecutive pages, as more than 200 in one place and as a specific count in another. This article says more than 200 and no more, and we are reconciling it.

Whether a reader can tell if this company is a peer. The published case page for this engagement carries no revenue scale. The scale used throughout this article, €1 billion, comes from our internal record. A case page that does not let a Director place the company is doing half its job, and we are fixing that page.

Where every figure above comes from

From the published case study page. The 97% service level designed, and the 10% waste reduction potential, both with the qualifier wording reproduced exactly as published. The consolidation of three warehouses into a single regional hub, the fragmented forecasting, the limited visibility across sites and the wastage from short shelf life products.

From our internal engagement record. The €1 billion revenue scale, the count of more than 200 product lines, more than 10 stakeholders engaged, more than five workshops, more than 10 scenarios mapped, the two analytical tools, and the 70% reduction in analysis time. The client’s Head of Supply Chain described the result as structure and clarity brought to inventory planning across regions.

No figure from the capacity block of our internal template appears anywhere in this article. That block is boilerplate, it is printed on most pages regardless of engagement, and it is not a measurement.

Clients are described by revenue scale and sector, never named.