Pricing and commerce
Product matching
Identifying the same product across retailers or sellers, using identifiers, attributes, text and images.
In practice
Every price comparison depends on it. A wrong match is a wrong price.
How matching works
A matching workflow links a retailer listing to a reference product. Start with an exact identifier where available, then compare model, size, colour, pack quantity and other distinguishing attributes. Similar titles alone are not enough: two listings can describe the same product family while selling different variants. Keep exact matches separate from comparable alternatives so downstream users know what is being compared.
Example: the same model, a different offer
Imagine two listings for the same headphones. One is a new single unit; the other is a refurbished bundle with a case. A title match could connect them, but using their prices as interchangeable observations would distort the comparison. Store the match decision, condition, bundle contents and supporting attributes alongside the source URL. This makes the decision reviewable rather than burying it in an unexplained score.
What to validate
Review a labelled sample of accepted and rejected pairs. Measure false matches as well as missed matches, and set a review queue for ambiguous cases. A high confidence score is not a guarantee of correctness. Recheck mappings when retailers reuse URLs, change variants or replace product content. Price changes should reflect market changes, not an accidental change in the identity of the item being tracked.
Aperture tracks prices, promotions and availability with AI product matching and dated evidence, and Import.io feeds deliver the same data to your own systems.
Aperture pricing intelligence →