Supply chain segmentation: running several chains at once

Segment on what changes the policy

Useful dimensions are those with different operational consequences: predictability of demand, contribution, required response speed, life-cycle stage, and whether the item is critical to the customer's own operation. Segmenting by product category or by sales region is administratively convenient and usually useless, because those groupings contain items that behave in completely different ways and therefore need different treatment.

Typical segment shapes

Predictable, high-volume lines suit efficient replenishment: larger order quantities, distant low-cost sourcing, stock held to a modest service target. Volatile or short-life lines suit responsiveness: shorter supply lines, later commitment, smaller and more frequent replenishment. Critical, low-volume items suit high availability regardless of turnover, because a shortage stops something expensive. The long tail suits a deliberately minimal policy, ordered on demand and reviewed for retirement.

Every segment needs a full policy set

A segment definition without policies attached is a classification exercise. Each segment should specify where it is sourced, what service level it targets, which replenishment model it uses, how it is planned, where its stock sits and how it is delivered. Writing those out side by side also exposes contradictions, such as a segment promised rapid delivery while sourced from the longest supply line available.

Sustaining the difference

Segmentation erodes because exceptions accumulate: a tail item gets a high service target after one complaint, a volatile line is sourced distantly to capture a price, and within a couple of years everything is being managed the same way again. Sustaining it requires periodic reclassification, a defined route for exceptions, and reporting that shows performance by segment so drift is visible. It also requires commercial agreement, since customers experience the differences directly.

Frequently asked questions

How many segments are workable?

Few enough that each has a genuinely different policy and that people can remember which is which. A handful is usually the practical limit, since every additional segment multiplies the parameter sets, the exception handling and the training required to operate them.

Should customers be segmented as well as products?

Often yes, because the same product can be served differently depending on the customer's contract, volume and service expectation. The combination produces a matrix rather than a list, and it should be kept coarse enough to remain operable.

Does segmentation conflict with a single service promise to the market?

Not if the promise is defined per proposition rather than universally. Problems arise when sales commits to a uniform standard while operations runs differentiated policies, which produces a promise the network was never designed to keep.