Speak one product language
Make every supplier and every channel speak one product language
Fuse converts source values into the reference lists your business uses for units, colours, sizes, materials, dimensions and technical characteristics.
The situation today
What goes wrong today
The same fact arrives written five ways. Noir, NOIR, Black, black. 1,5 kg, 1500 g, 1.5 kilogram. Coton 100%, 100 % coton, pure cotton. M, Medium, taille intermédiaire. Bleu marine, Navy, Marine. Every spelling becomes its own facet value on the shop, so filters fragment, comparison between two products stops working and the search result depends on which supplier loaded the row.
The mechanism
How it runs in Fuse
You define the target list once. Harmonizer applies deterministic cleanup on the way in: whitespace, number and date formats, GTIN shape, unit aliases and the glossary terms you approved. Mapping and glossary carry colour, size and material onto your own values. Where you have a taxonomy, Classifier proposes the category and says when it is unsure. Values it cannot place are reported as unknown rather than guessed, so the list grows by decision instead of by accident.
Fuse does not arrive with a ready-made value list for every attribute. Colour, size, material and category come from your mapping, your glossary and, where a taxonomy exists, Classifier. Validation detects remaining problems; it does not rewrite the record.
What changes
What you get
-
Filters that behave
One value per concept, so facets and search return what the shopper expects to see.
-
Comparable products
Units and dimensions in the same shape across every supplier, so specifications line up.
-
Fewer keying errors
The reference list is the reference. Free text stops setting the standard by default.
-
Rules in one place
No per-channel spreadsheet of substitutions to maintain and re-synchronise.
Indicators to track
How to measure it
- Values normalised
-
values mapped per import
- Attributes matching the reference list
-
percentage of attributes in scope
- Unknown or ambiguous values
-
count per import, listed for review
This number should fall as the reference lists absorb real supplier vocabulary.
- Products correctly filterable
-
percentage of the catalogue with complete facet attributes
Next
Related use cases
-
Supplier onboarding
Fuse reads the file or the API, infers its structure, proposes the mapping to your model and flags blocking errors before they reach the ERP, the PIM or the shop.
-
Catalogue localisation
Titles, descriptions, attributes and categories move into the locales you configure, with local units, your own vocabulary and every untranslated field flagged.
-
Marketplace feeds
From one golden record, Fuse builds the feed each channel expects and reports the non-conformities, product by product, before publication.
Show us the attribute your filters keep fragmenting
Colour, size, material or a unit of measure. We map the incoming values onto your reference list and show you what is left for a person to decide.