SKU rationalisation on autopilot:
know which products actually earn their shelf
A catalogue grows for years and rarely shrinks, because cutting a SKU feels riskier than keeping it even when nobody can say why it is still there. We build a model that ranks every SKU by real margin, sales velocity and carrying cost, so a cut, keep or relaunch decision is based on numbers instead of habit.
The process today
Most catalogues grow steadily and prune rarely, because adding a SKU is an easy decision made in the moment and cutting one requires someone to actively justify removing something that already exists, with no clear owner for that review. Over a few years, a catalogue accumulates SKUs that sell a handful of units a month, tie up warehouse space, and nobody can say with confidence whether they are worth keeping.
The cost is distributed and therefore invisible in any single number: a little extra carrying cost here, a little cannibalisation of a better SKU’s sales there, a little operational complexity in managing variants that barely move. None of it shows up as a dramatic problem on its own, but the cumulative effect on margin and operational simplicity is real.
The review that would catch this, ranking every SKU by true profitability, not just revenue, factoring in carrying cost and the risk of cannibalising a better seller, is tedious enough that it rarely gets done properly, and when it does happen it is usually a one-off project rather than something revisited on any regular schedule.
What the agent does
The model ranks every SKU by true margin, sales velocity and carrying cost together, rather than any one metric in isolation, and produces a cut, keep, relaunch or watch recommendation with the specific numbers behind it. A SKU that looks weak on pure sales volume but has a strong margin and low carrying cost can rank fine; a SKU with decent volume but thin margin and high carrying cost, aging stock risk, storage space, can rank for review even though its top-line sales look unremarkable.
Before recommending a cut, the model checks for cannibalisation: how much of that SKU’s demand would likely transfer to a similar SKU you keep versus be lost outright, so the decision accounts for what cutting it would actually do to total revenue, not just that one line’s revenue. Seasonal SKUs are flagged specifically, so a product that is genuinely slow right now but strong in its season three months from now does not get recommended for a cut based on a snapshot that caught it at the wrong time of year.
The dashboard lets a merchandising team view the ranking by category, brand or supplier, so a review can happen at whatever level makes sense for a given decision, a single SKU or an entire underperforming line from one supplier.
What stays with humans
The actual cut, keep or relaunch decision stays with your merchandising team, along with any judgment about strategic SKUs kept for reasons the model cannot see, a loss-leader, a flagship product, a supplier relationship worth preserving. The model prepares the ranking and the reasoning; it does not remove a product from the catalogue on its own.
Guards
Every recommendation is logged with the margin, velocity and carrying cost numbers behind it, so a merchandising review can be audited later. The model’s judgment is checked against SKUs you have already discontinued in the past, to see whether its recommendation would have matched what you actually decided, and seasonal flagging prevents a premature cut on a product that is simply between its selling seasons.
Price and timeline
| Option | Price | What it covers | Timeline |
|---|---|---|---|
| Single automation | from $900 | Full catalogue ranking, cut/keep/relaunch recommendations | 7 to 14 days |
| Department package | from $2,800 | SKU rationalisation plus demand forecasting on the surviving catalogue | 3 to 5 weeks |
Running cost is usually $20 to $70 a month depending on catalogue size.
Related
Pair this with demand forecasting so the SKUs you keep get a real forecast, not a guess, and with price elasticity analysis so a borderline SKU’s price, not just its existence, gets reviewed too. The full package breakdown is on the AI agents service page and the automation-everything overview; for real catalogue work, see the Thailand D2C rebuild case study and the Balkans supplements store case study.
Ready to see which SKUs are actually earning their shelf space? Get in touch and we will look at your catalogue in the first call.
Tired of doing this by hand? We can take the whole routine off your team, not just this step: Routine takeover, from $400 →
FAQ
How much does SKU rationalisation automation cost?
From $900 for a ranking and recommendation pass across your catalogue, live in 7 to 14 days. A department package adding demand forecasting on the surviving catalogue usually starts at $2,800.
Will it just tell us to cut everything that sells slowly?
No. A slow-selling SKU that is highly seasonal, strategically important, or genuinely low-cost to carry can score fine on this model. The recommendation accounts for carrying cost and margin together, not sales velocity alone.
What is cannibalisation checking for?
Before recommending a cut, the model estimates how much of that SKU's demand would likely shift to a similar SKU you keep versus be lost entirely, so a cut decision is not made blind to what it would actually do to total revenue.
What data does it need?
Sales history, cost and margin data per SKU, and carrying cost estimates (storage, aging, markdown risk) if you track them. Where carrying cost is not tracked precisely, we use a reasonable estimate and flag it as such.
Does this replace a human merchandising review?
No, it prepares the data for one. The model's ranking and reasoning make a merchandising review faster and better-informed; the actual cut decision is a business call your team makes.