Segmentation Analysis 2026: Where the Margin Hides

A customer segmentation is a hypothesis about cost

Cutting a customer base produces buckets. Buckets become a segmentation only when the groups differ in something you can act on. Three tests decide whether they do.

1. Do the groups have different economics? Different contribution margin, different cost to serve, different retention. Same economics means one segment wearing two labels.
2. Does a different action apply to each? If the response to every group is the same price change, the cut has not divided the problem.
3. Can you tell which group a customer is in before you serve them? A segmentation you can only apply in hindsight cannot be operated.

Test three is the one that kills elegant cuts. Segmenting by profitability is circular unless something observable predicts it. That observable thing is usually what the cut should have used in the first place.

Simon-Kucher, whose published focus is pricing and commercialization, lists structured problem solving and the ability to prioritize key drivers among what it watches, checked 30 September 2026 at simon-kucher.com. Prioritizing a driver is exactly what turns a bucket list into a segmentation.

The contractor, cut by contract size

A commercial cleaning contractor turns over 52 million dollars a year and reports 6.5 percent operating margin, which is 3.38 million dollars. It runs 402 sites under three contract bands.

Contribution adds to 6.2 plus 2.5 plus 0.3, which is 9.0 million dollars. Overhead is 5.62 million, so operating profit is 9.0 minus 5.62, which is 3.38 million, and it closes on the reported figure.

The small band already looks weak at 4.8 percent contribution margin against 19.9 percent for large sites. What the table does not yet show is how much overhead each band consumes, and that is where the answer is. Small sites are 6.2 of 52 million dollars, which is 11.9 percent of revenue, and 214 of 402 sites, which is 53.2 percent of the estate.

Profitability analysis by driver rather than by revenue

Of the 5.62 million dollars of overhead, 3.9 million is area supervision: regional managers who visit sites, audit quality and handle client complaints. That cost is driven by site count, because a visit takes the same time whatever the contract is worth. The remaining 1.72 million is finance, payroll and insurance, which does not scale with sites.

Supervision per site is 3,900,000 divided by 402, which is 9,701. Round it to 9,700 dollars a site and allocate.

The three results sum to 5,831,400 plus 1,045,000 less 1,775,800, which is 5,100,600 dollars, and that equals 9,000,000 less 3,899,400. The allocation closes.

Small sites lose 1,775,800 dollars a year after the supervision they cause. That is 52.5 percent of the company's entire operating profit of 3.38 million dollars. Three earlier views all missed it. The average margin of 6.5 percent missed it. The revenue split, where the band is 11.9 percent, missed it. The contribution table, where the band still showed a positive 300,000 dollars, missed it too.

Try the other base. Allocated by revenue, small sites absorb 11.9 percent of 3.9 million, which is 465,000 dollars. The band would then show a loss of 165,000 dollars. The two bases differ by roughly a factor of ten on the same segment. Only one of them describes what causes the cost. Charles River Associates says it watches "what data you draw on", checked 30 September 2026 at crai.com. The allocation base is that choice in its purest form.

What exiting the segment is worth

The obvious move is to exit the small band, and the arithmetic of the obvious move needs one more step before it is a recommendation.

Exiting gives up 300,000 dollars of contribution and removes 214 sites from the supervision load. If the supervision cost actually leaves, the gain is 2,075,800 minus 300,000, which is 1,775,800 dollars, and operating profit rises from 3.38 million to 5.16 million.

Supervision is a step cost. At roughly one supervisor per 22 sites, 214 sites carry 214 divided by 22, which is 9.7 supervisors. Exiting all 214 removes nine or ten roles. It removes them only if the remaining 188 sites regroup into whole territories. A contractor that exits the sites and keeps eleven supervisors has given up the contribution and kept the cost.

So the recommendation has a sequence in it. First, reprice. Average revenue per small site is 6.2 million divided by 214, which is 28,972 dollars. Supervision is 9,700 dollars against that, so it is a third of the revenue of a small contract. A price rise of a third is what makes the band viable, and it is not obviously achievable. Second, exit the sites that refuse, in whole territories rather than one by one. Third, take the supervisor headcount out as each territory clears.

Customer segmentation cuts that describe and cuts that decide

Contract size worked here because it tracks the cost driver. Most candidate cuts do not. The failure is quiet, because any cut at all produces a table.

Deloitte's Recreation Unlimited case publishes a client with 5 percent of sales online against a sector average of 10 percent, checked 30 September 2026 at usrecruiting.deloitte.com. That is a channel cut with a gap attached. It works because the two channels carry different fulfillment costs, different margins and different growth rates. A cut by customer age would produce a table on the same data and support no decision at all.

Three cuts that usually decide something, in services businesses like this one.

1. Cost to serve. Visits, calls, deliveries or tickets per account, which is the driver that produced the finding above.
2. Contract structure. Fixed-price against cost-plus, because they transfer risk differently and respond to inflation differently.
3. Tenure. First-year accounts usually cost more to serve and price lower, and separating them stops the average from hiding both.

Simon-Kucher publishes a claim that its projects raise client profitability by 100 to 500 basis points, checked 30 September 2026 at simon-kucher.com. Basis points of profitability are what a segment-level repricing moves, and on this contractor 1.78 million dollars against 52 million of revenue is 342 basis points.

Where segment work is scored

The weights below are ours, read from the rubric code our own grader runs. No consulting firm publishes a marking scheme, so none of this is Simon-Kucher's view of a segmentation.