Product matching: definition, methods and target rates

Product matching involves associating each item in your catalog with its exact equivalent at your competitors' stores, to compare prices that truly reflect the same product. Two products are equivalent if they share the brand, model, variant, and packaging: a six-pack of bottles is not comparable to a single bottle. This is a difficult task because each retailer writes its descriptions differently and doesn't always display the EAN code. This is the foundation of any price monitoring: an incorrect match leads to a false price difference, and therefore a poor decision.

Automated product matching proceeds in successive passes, from the most reliable to the least reliable. It first matches products that share the same EAN code . It then compares standardized attributes (brand, manufacturer reference, capacity, color), and finally analyzes labels and images using natural language processing. Each match receives a confidence score: reliable matches are automatically validated, while uncertain matches undergo human review. To apply this method to your competitive intelligence, see how to improve the reliability of competitive product matching .

Exact matching pairs two identical items, while equivalent product matching pairs different but comparable items for the customer. The former applies to branded products sold everywhere under the same EAN code. The latter is used to compare products without strict equivalents, such as store brands: a 500g package of store-brand pasta is compared to competing store brands of the same size and similar quality. This second case requires explicit similarity rules and, often, a price adjusted per unit of measure to correct for differences in size. This is also known as product linking .

Yes, product matching needs to be redone regularly, as catalogs are constantly evolving: new products, discontinued lines, packaging changes, and descriptions modified by competitors. A valid match today can become invalid when a competitor replaces a 1-liter format with a 900 ml one without changing the main description. Therefore, we combine continuous matching of new products with a periodic review of existing matches, more frequent for closely monitored products like KVIs. A useful indicator is the proportion of matches that are called into question with each review: if it increases, the rules need to be revised.

Product matching can be almost entirely automated, but some human review remains useful. High-confidence matches, such as those with the same EAN code, are validated without intervention. Ambiguous cases (similar variants, pack sizes, products without EANs, private label brands) benefit from validation by someone familiar with the category. Each human validation also helps improve the algorithm. The realistic goal is to reserve human intervention for matches where an error would be most costly, rather than aiming for complete automation.

Product matching software is judged first and foremost by the reliability of its matches, measured on a sample of your own catalog, and not just by its coverage rate. Verify that it combines multiple signals (EANs, attributes, labels, images), displays a confidence score for each match, and allows for easy error correction. Also consider its ability to handle equivalent products, such as private label brands, and to track catalog changes over time. Booper offers product matching software that assigns a confidence score to each match.

Product matching brings your product references closer to those of your competitors, while deduplication brings together references in the same catalog that actually refer to the same item. Internal duplicates appear when a product is created twice, for example, on two sales channels, or after the merger of two brands. They scatter sales history, distort elasticity and forecasts, and complicate competitive matching. The techniques are similar (EAN codes, attributes, labels), but the stakes differ: see how to deduplicate your product catalog .