Why Your SKU Portfolio Is Quietly Killing Margin

Almost every manufacturer I speak to has run a SKU rationalisation exercise at some point. Almost none of them can tell me what it achieved. The tail got trimmed, a few dozen codes were retired, everyone moved on, and eighteen months later the portfolio had grown back to roughly where it started, with a fresh set of low-volume variants nobody can quite account for.

That pattern is not a failure of discipline. It is a failure of framing. SKU rationalisation is treated as a project: run once, by a small team, in a spreadsheet. But portfolio complexity is not a project-shaped problem. It accretes continuously, driven by sales commitments, customer-specific packs, promotional variants and trade demands. A one-off cut against a continuously growing problem is, arithmetically, a losing strategy.

The spreadsheet is doing more damage than the tail

The standard exercise pulls twelve months of volume and revenue by SKU into Excel, sorts descending, and draws a line somewhere around the point where the cumulative curve flattens. Everything below the line becomes a candidate for deletion. It feels rigorous because it is quantitative. It is not rigorous, because it is quantitative about the wrong things.

Volume and revenue are the two variables least likely to tell you whether a SKU is worth keeping. The variables that matter are mostly absent from the analysis:

• True cost to serve, changeover time on the line, minimum batch economics, the specific packaging format, cold-chain or handling exceptions.
• Strategic coupling, whether a low-volume SKU is the reason a high-volume listing exists at all with a given retailer.
• Cannibalisation, whether deleting it moves volume to a sibling SKU or simply loses it.
• Working capital drag, slow-moving finished goods, plus raw materials and packaging held only for that variant.
• Forecast error, the SKUs that consume disproportionate planner attention and generate the most write-off.

None of this is exotic data. It sits in the ERP, the MES, the WMS and the trade spend system. It is simply not assembled, because assembling it by hand for two thousand SKUs across five plants is a multi-week analyst exercise nobody can repeat quarterly. So the analysis defaults to the two fields that are easy to pull, and the decision defaults to the tail.

Why the cuts don't hold

Even where a cut is well-reasoned, it usually doesn't survive contact with the commercial organisation. Sales pushes back on a specific deletion, citing a customer relationship. There is no shared, defensible basis on which to arbitrate, because the deletion logic lived in one analyst's workbook while the counter-argument lives in a regional manager's account knowledge. The loudest advocate wins, the exception is granted, and the exception becomes the precedent.

A rationalisation decision that can't be re-run next quarter isn't a decision. It's an opinion with a spreadsheet attached.

Treat it as a scoring problem, not a cutting exercise

The alternative is to stop running rationalisation as an event and start running it as a standing capability. Three things change.

1. 1Every SKU carries a live composite score. Margin contribution after true cost to serve, volume trend, forecast accuracy, working capital consumption, strategic coupling and cannibalisation risk are computed continuously from data you already produce, not assembled by hand at review time.
2. 2Deletion candidates are proposed, not decreed. The system surfaces a ranked list with the reasoning attached to each line: this SKU scores poorly because changeover cost per unit is four times the category average and its forecast error is 60%. Sales can see why, and argue against a specific input rather than the conclusion.
3. 3Every proposed cut is simulated before it's taken. What-if analysis models the volume that transfers to sibling SKUs versus the volume genuinely lost, the freed line capacity, the working capital released and the effect on the customer's listing. The decision is made against a modelled outcome, not a hope.

Make it a quarterly forum, not an annual project

The structural fix is governance. A standing quarterly portfolio review, attended by supply chain, finance and sales, working from the same live scorecard, with a documented decision on every candidate, keep, delete, reformulate, or reprice. Exceptions are granted with a stated reason and a review date, not permanently.

That cadence does something a one-off cut can never do: it catches proliferation on the way in. A new customer-specific variant gets scored at introduction against the same criteria, and someone has to argue for it on the record. Most of the portfolio bloat I see was never decided, it was simply never questioned.

The tail isn't the problem. The absence of a repeatable, shared, evidence-backed process for questioning the tail is the problem. Fix that and the tail manages itself.