Photo: Google Gemini · AI-generated
Here is a sentence from a real project that explains the entire problem:
Stock appeared adequate at the item level while specific sizes were out of stock.
That is apparel in one line. On the dashboard, the product looks healthy. On the shelf, the sizes people actually buy are gone. Nothing is technically wrong with the system — it is answering a different question than the business is asking.
One product is not one thing
Most retail software is built around a product record. A product has a price, a stock level, a supplier. That model works for a kettle.
It does not work for a garment. One style, in five colours, across seven sizes, is thirty-five separate things that must be counted, reordered, and reconciled separately. They share a name and almost nothing else. Demand is uneven and predictable in shape but not in detail: mid sizes move, the extremes do not, and one colour outsells the rest for reasons nobody forecast.
Systems that treat the style as the unit will happily report health while the sellable combinations are gone. The number on the screen is real. It is just not the number that decides whether you can fulfil the next order.
The reconciliation problem underneath
The variant matrix gets harder when it spans systems, which it almost always does.
We built a supply-chain intelligence platform for a garment manufacturer selling on Amazon. They ran two capable systems that did not talk to each other: Zoho Inventory for procurement, stock, and fulfilment; Sellerboard for Amazon sales, fees, advertising spend, and profitability. Both did their jobs. Neither could answer management’s actual questions.
The symptoms are worth listing, because they are the same everywhere in this sector:
- Stock looked fine at item level while individual sizes were unavailable
- Sales figures differed between systems with no automated way to explain the gap
- Profitability used manually entered costs that missed the full landed cost of an imported garment
- Every reorder decision meant reconciling warehouse quantities, Amazon FBA levels, open purchase orders, and sales velocity by hand, across spreadsheets
None of that is a missing feature. It is a missing layer — something that reconciles two systems at variant level and produces one consolidated view. That is integration work, and it is where most of the real difficulty in fashion retail software lives.
Landed cost is a variant-level number too
The profitability point deserves its own paragraph, because it is the one most often waved through. For an imported garment, unit cost is not the invoice price. It is the invoice price plus freight, duty, and handling — and those allocate differently across variants. A system that stores one cost per style will report margins that are confidently wrong, and the error is invisible precisely because the number looks so clean.
If you cannot see true margin per variant, you cannot tell which colours to reorder and which to discontinue. That is not a reporting inconvenience. It is a buying decision made with the wrong information, every season.
The other extremes: one-of-one, and 500 screens
Apparel commerce does not only stretch software at the variant matrix.
At one end sits resale, where the matrix collapses entirely: on Atlikarinca, a consumer marketplace, every listing is a single unique item with its own condition, photos, and negotiated price. There is no stock level to speak of — the data problem becomes trust, ratings, and matching instead.
At the other end sits physical retail at scale. For a jewelry manufacturer in Hessen, we rebuilt the in-store kiosk application running across more than 500 touchpoints in Europe, on native iPad, against SAP data through an existing middleware layer. Here the challenge is consistency: the same catalogue, correct everywhere, on hundreds of screens, coordinated across an in-house SAP team and a middleware vendor.
Same industry. Three completely different shapes of problem.
What this means when you are buying software
The point is: in fashion and apparel, the variant is the unit of truth. Any system that treats the style as the unit will eventually mislead you — and it will do so quietly, in a dashboard that looks fine.
So the questions worth asking before you commit to a platform are narrow and specific. Can it hold stock, cost, and margin per variant, not per style? Can it reconcile those variants across every system you actually run? And when it cannot, is there a layer that can be built to close the gap — rather than a spreadsheet and someone’s Friday afternoon?
If your stock reports look healthy and your best sizes keep selling out, we should talk.