Product matching for price comparison: how it works

· 3 min read · CrawlMine guides

Why matching matters

A competitor price is only useful when you know which of your products it belongs to. Stores describe the same item differently, sell it in different pack sizes, bundle it with other things or list several variants on one page. If matching is wrong, every comparison built on it is wrong too.

Matching is the least visible part of a price monitoring project and often the most valuable.

Step 1: Use identifiers when they exist

The cleanest match is an exact shared identifier:

  • GTIN, EAN or UPC barcodes, shown on some product pages
  • Manufacturer part numbers (MPN) and model numbers
  • Marketplace IDs, where you already know your own listing

When a competitor shows an identifier that matches yours, the match is near certain. But many stores do not show them, and some show them inconsistently.

Step 2: Normalise the listing

Before comparing text, clean it up:

  • Lower-case and remove punctuation and filler words
  • Extract brand, model, variant (size, colour, capacity) and pack size
  • Convert units: 1 kg and 1000 g, 6.1 inch and 15.5 cm
  • Detect bundles and multi-packs

Normalisation turns “Brand A Trail Runner 2 Men’s Running Shoe, Blue, UK 9” into structured attributes that can be compared field by field.

Step 3: Find candidates

For each of your products, find a short list of competitor listings that could be the same item: same brand, similar model, same category. This keeps the comparison focused and fast.

Step 4: Score each candidate

Compare your product with each candidate on brand, model, variant, pack size and title similarity. Combine them into a confidence score. Exact brand and model with matching size scores high. Same brand and similar title with a different pack size scores lower.

Step 5: Decide with thresholds and people

  • High confidence: accept automatically.
  • Medium: send to a person to confirm.
  • Low: treat as no match.

The person reviewing only sees the uncertain cases, and only once. Their decision is stored and reused on every later run. Over a few weeks, the review queue shrinks to new products only.

Pack sizes and unit prices

Often the same product is sold in different sizes. A 500 g pack and a 1 kg pack are not the same product, but comparing their unit prices is useful. Good matching records both: whether it is an exact match, and a unit price for like-for-like comparison. This is essential in grocery and quick commerce.

Variants

A page may show one product with many sizes or colours, each with its own price and stock. Matching should work at the variant level. A price for size S says nothing about size XL.

Bundles and kits

“Phone with case” is not “phone”. Detect bundles and either exclude them or record them separately with their contents, so they do not distort price comparisons.

Measuring match quality

Ask for a match report:

  • How many of your products were matched per competitor
  • The confidence distribution
  • How many were reviewed by a person
  • A sample of matches you can check yourself

Unmatched products are not always a failure. Sometimes the competitor simply does not sell the item, which is useful to know too.

Where this is used

Product matching sits under e-commerce price monitoring, MAP monitoring and product catalogue data. If you want to see match quality on your own catalogue, send a product list and competitor sites or email [email protected].

Talk to us directly

[email protected]

Send the websites and fields you need. A person on our team reads every message.

Email us

Send us the sites tonight. Data by morning.

Tell us which websites and fields you need. We confirm what is practical, how it should be delivered and what it costs.

  • Goes straight to [email protected]
  • Reply within one working day
  • A sample or pilot before any recurring plan
  • NDA available before you share details

[email protected]