Parent / variant: definition

Why it matters

• Cost: enriching every variant as an isolated product means paying for the same work several times. Enriching the model and then applying the result reduces variant processing to a few seconds instead of a complete analysis.
• Consistency: two variants of the same model must carry the same shared specifications. Independent processing always diverges eventually; grouping makes consistency structural.
• Control: collateral completion — a value established at model level propagates to its variations — creates harmony and turns any divergence between variants of the same model into a detectable error signal.

This is why grouping is the first step in the chain: categorisation, enrichment and matching all benefit from it. A group is defined by its variation axes — size, colour and sometimes several combined axes — and can be reconstructed even when the catalogue carries no explicit structure.

When the catalogue is flat

Many catalogues have no explicit hierarchy, so it has to be reconstructed. The robust method works on sorted product blocks. Looking at a few dozen products at once reveals series — one model followed by products whose only differences are size and colour — whereas product-by-product processing is short-sighted and splits families in two. One question decides whether a difference between adjacent products is a variation axis: “if these two products differed only in this attribute, would the rest of their technical data be identical?” If yes, it is a variation axis, such as a light bulb’s colour temperature or a cable’s length. If not, it is a model-defining attribute, such as power or cross-section. Variation axes are discovered family by family; assuming they are always “size + colour” is the classic mistake.

Doubt is asymmetric: when uncertain, split the group. Excessive splitting has little consequence; excessive merging propagates false values between products that are not truly siblings. An honest limitation is that ranges made up of purely technical references, without a “name + variation” logic, must be grouped cautiously. Without the manufacturer’s nomenclature, there is no reliable structure to guess.

Grouping is also checked against field evidence. Distributor product pages often list all variations of a model themselves through size and colour selectors. When a third-party page groups the same products, the grouping is confirmed; when it lists others, the relationship is enriched.

In the PIM

Modern PIMs carry this structure natively. In Akeneo, family variants define which attributes belong to the variant; any attribute not listed belongs to the parent and is shared. On delivery, shared attributes are written once on the model and specific attributes on each variant. See also: product data enrichment · Akeneo integration.