The system proposed the order. Everyone checked it by hand anyway.

The system proposed the order. Everyone checked it by hand anyway.

An 8 to 10 week sprint. A €1B Nordic consumer retail group. Part three of three on where supply chain work stalls. Part one was a decision rule nobody had written down. Part two was an owner nobody had named. This one is a system nobody trusted.

The client runs a central warehouse that pushes product to more than 350 stores across more than 700 product lines. The system generated order proposals. That is what it is for.

Nobody believed them. The page’s own wording is that planning relied heavily on manual checks and fragmented data across systems, and that this reduced confidence in system-generated order proposals.

Read that chain in the right direction, because it is easy to read backwards. The data was fragmented first. The manual checking came second, as a rational response. The loss of confidence came third, and by then it was justified. Nobody was being difficult. Every planner overriding a proposal by hand had been right often enough to keep doing it.

That is the specific trap. A system that is wrong often enough to need checking will be checked forever, and the checking is what removes the time you would need to fix the data underneath it.

The people who could fix it were the people the checking was already consuming.

The page names the barrier in one sentence and it is the whole reason this engagement existed: purchasing teams were at full capacity managing daily operations, leaving little time to analyze data or improve forecasting accuracy.

That sentence is worth slowing down on, because it is not a complaint about headcount. It is a description of a loop. The manual checking exists because the forecast is not trusted. Improving the forecast requires the people doing the manual checking. They are not available, because they are checking.

Nothing about that loop is anybody’s fault, and nothing about it resolves on its own. It is stable. It will run for years, at a cost that never appears as a line item, because the cost is a team’s attention rather than a payment.

There is a second reason, quieter than the first. Segmenting 700 product lines by how predictable their demand is takes a decision about where the boundaries go, and that decision is easier to postpone than to make. A fast, predictable line and a slow, erratic one are obvious. The middle is not, and the middle is where most of the lines are.

This is the general shape of a trust problem in planning. What is specific to this client is on the page: fragmented data across systems, heavy manual checks, and confidence in the proposals falling as a result.

Two figures, both correctly worded, and that is worth saying out loud

Both percentages carry the words potential identified. Neither is presented as a delivered saving.

We have spent two of the last three pieces pointing out that our own case pages state modelled figures as achieved results. It would be dishonest not to say the opposite when it is true. This page is worded the way the whole estate should be, and so was the out-of-box failures page. Two in a row is not yet a habit, but it is not an accident either.

What the sprint delivered is a segmentation, rebuilt replenishment quantities and tools in the planners’ hands. Whether 30% of the manual effort actually disappears depends on whether the tools are used and whether the data underneath them stays clean. That is the honest reading and the page supports it.

Ten interviews is not the expensive part. Sorting 700 lines is.

The page reports 10+ interviews conducted and 3 processes mapped. That is a modest fieldwork number by the standards of these sprints, and it is the right one here, because this engagement was not a discovery problem. Everybody knew the proposals were being overridden. What nobody had was the segmentation.

Sorting more than 700 product lines by sales and by demand predictability is not a meeting. It is a sustained piece of work on messy data that lives in more than one system, and the reason it had not been done is in the section above: the only people who understood the lines well enough to sort them were the people whose days were spent checking orders.

That is the whole trade, and it is worth being exact about it rather than hinting. The client knew their stores, their products and their seasonality far better than any outside team will learn in ten weeks. None of that was missing. What was missing was somebody whose entire ten weeks could go into the sort, and who had done the sort before.

Because that is the part that repeats. A retail planning team builds a demand segmentation once, maybe twice, in a career, and the first one is always the slow one. It is not a hard technique. It is a technique with a hundred small decisions in it, and knowing which ones matter is worth more than knowing the method.

Fix the trust, not the checking

The instinct in a situation like this is to attack the manual work directly, because the manual work is what everyone can see and what everyone resents. That is the wrong end.

The checking is not the problem. The checking is the symptom, and it is a rational one. Remove the checks while the proposals are still unreliable and you have not saved effort, you have moved the cost from a planner’s afternoon to a store’s empty shelf.

The order that works is the one the sprint used. Make the proposals worth believing, by segmenting and rebuilding the quantities underneath them. Then the checking falls away on its own, because a planner who has watched the proposals be right for a month stops opening them all.

That is also why the tools mattered more than the analysis. A segmentation degrades. New products arrive, demand patterns move, and a sort that is two years old is quietly wrong in the middle where it was always hardest. A tool the team can re-run is the difference between a fix and a fix with a shelf life.

Three things, and none of them is the number

Whether the checking actually stopped. The 30% is a potential identified against the manual effort in the process as mapped. It becomes real only when planners stop opening proposals they no longer need to open, and that is a habit rather than a system setting. Habits outlive the reason they were formed.

Where the segmentation boundaries should sit in the middle. The fast and predictable lines sort themselves. The slow and erratic ones sort themselves. The middle band is a judgement, it was made with the data available in that window, and a different and equally defensible boundary would move a meaningful number of lines. The top and bottom of a segmentation are robust. The middle is where reasonable people differ.

Whether the data stays clean. The whole result rests on the data being less fragmented than it was. Fragmentation is not an event that happened once, it is a condition that returns whenever a new system is added and nobody owns the join. This sprint did not, and could not, settle who owns that.

Where every figure above comes from

From the published case study page. The €1B Nordic consumer retail group descriptor, the 8 to 10 week sprint, the central warehouse serving 350+ stores across 700+ product lines, the 10+ interviews conducted and 3 processes mapped. The 30% reduction in manual planning effort and the 20% improvement in stock efficiency, both carried here with the page’s own qualifier of potential identified, and the 15% shortening of the replenishment cycle, which the page reports inside the 20% line. The statement that planning relied heavily on manual checks and fragmented data across systems, reducing confidence in system-generated order proposals. The statement that purchasing teams were at full capacity managing daily operations, leaving little time to analyze data or improve forecasting accuracy. The three objectives set for the sprint. The client’s Head of Supply Chain Coordination described the outcome as clarity and tools that made replenishment planning faster and more reliable.

Deliberately not used. The segmentation itself, where the boundaries were drawn, and which product lines sat in which band. That is the client’s commercial detail, it is the most directly useful thing in the engagement to a competitor, and it is not ours to publish. Every argument here is about the method that produced the segmentation, which is the part that transfers.

Not claimed anywhere. That manual planning effort fell by 30%, or that stock efficiency improved by 20%. Those are potentials identified. We handed over a segmentation, rebuilt quantities and tools. What happened next is theirs.