Stock Keeping Unit

sku · stock keeping unit · stock-keeping unit · article number · product variant · item code

Definition

An SKU is the unique identifier of one product variant, defined by colour, size or configuration; it is the primary unit of accounting in a catalogue and in recommendation systems.

SKUs and the catalogue in e-commerce

One physical product in an online store usually has several SKUs. A Nike Air Max 90 in four colours
and seven sizes is 28 SKUs. Each SKU carries its own price, or price range, its own stock level,
its own photographs and its own specification.

For recommendation algorithms, the SKU is the object the shopper interacts with. A view event is
attached to one specific SKU, so the algorithm does not merely know that someone looked at
trainers: it knows they looked at the Nike Air Max 90, in white, size 43.

SKU versus product: which level to collect events at

Level	When to use it	Example
SKU	Fashion, footwear, clothing	The exact colour and size matter
Product (parent)	Electronics, books	The variant is secondary, the product is primary
Mixed	Most catalogues	SKU for inventory, product for recommendations

Different verticals need different strategies. In fashion, size is critical: recommending a product
that is out of stock in this shopper’s size costs conversion. In electronics, people choose the
model first and the configuration second.

SKUs and cold start

New SKUs start with no interaction history and suffer from the cold start problem: the
recommendation algorithm does not know who to show them to. The fixes are attribute similarity with
other SKUs, a boost through merchandising rules, and placement in new-arrivals blocks. Without
active management, a new SKU can receive almost no traffic from recommendations and take a long
time to accumulate organic data.

Frequently asked questions

How does an SKU differ from a product ID?

A product ID, sometimes called the parent product, identifies the product as a concept, such as the Nike Air Max 90. The SKU is one specific variant of it, such as the Nike Air Max 90 in white, size 42. One product ID contains several SKUs. In analytics and recommendations it matters which level the events are collected at.

Should recommendations be built at SKU or product level?

It depends on the vertical. In fashion and footwear, recommendations work at product level, because the shopper sees the product rather than one size. In electronics the SKU level is critical, since configurations differ substantially in price and specification. A recommendation system should support both levels.

What is catalogue chaos and where do SKUs come into it?

When one physical product carries several different SKUs across systems — the ERP, the website, the warehouse management system — recommendation algorithms cannot aggregate behavioural data correctly. Normalising SKUs is a precondition for personalization working well.

Gravity Field

SKUs in Gravity Field

Gravity Field processes events at SKU level and supports aggregation up to product level for recommendations and analytics. SKU data is synchronised through the product feed and used across every personalization module.

✦Event support at SKU level (view, add to cart, purchase)

✦Aggregation of behaviour up to the parent product where needed

✦Recommendation filtering by SKU attributes (price, availability, category)

✦Catalogue updates through the product feed with no personalization lag

Explore the platform →