cd /news/ai-products/ai-product-matching-for-tracking-com… · home › topics › ai-products › article
[ARTICLE · art-171776] src=blog.apify.com ↗ pub= topic=ai-products verified=true sentiment=↑ positive

AI product matching for tracking competitor prices

Apify's AI Product Matcher Actor uses a machine learning model to identify the same product across competing stores despite differing titles, demonstrated on 19 Hifiman listings from Headphones.com and 98 from Audio46 that produce 1,862 possible pairs. The tool takes two scraper-collected datasets, maps fields such as title, price and description through JSON attribute mappings, and outputs a predicted_match flag of 1 or 0 plus predicted_scores for each pair. Users choose a precision setting for more reliable matches or recall for broader but less accurate suggestions.

key takeaway

Mapping source fields like title to the matcher's expected name field via JSON attribute mappings lets the Actor compare two store datasets and flag matches with predicted_match (1 or 0) and predicted_scores, with a precision setting favoring reliable matches over recall.

by read6 min views4 publishedNov 18, 2026
AI product matching for tracking competitor prices
Image: Blog (auto-discovered)

To price your products competitively, you need to know what other stores charge for them. But that can be tricky when competitors use different names and descriptions for the same item.

You’d need to manually check model numbers and specifications just to confirm a match. Doing that for a few listings is manageable, but the work piles up across hundreds or thousands of products.

AI Product Matcher on Apify helps automate that process.

What is AI product matching? #

AI product matching uses product details to identify the same item across different stores. For example, “Hifiman Sundara Closed-Back Headphones” and “Hifiman Sundara Closed-Back Planar Magnetic Headphones” describe the same model, despite their different titles.

After a scraper collects the listings, AI Product Matcher uses a machine learning model to find matching products across your datasets.

How to compare competitor prices with AI Product Matcher #

The example below uses 19 Hifiman listings from Headphones.com and 98 from Audio46. Comparing every listing in the first store with every listing in the second creates 1,862 possible pairs. The matcher processes those pairs and marks the ones it considers matches.

Step 1. Add the sample datasets

The product listings for this example were collected with Shopify Scraper from Apify Store. You can use the resulting datasets directly to try the matcher without scraping the data yourself.

Open AI Product Matcher, select Try for free, and sign in to Apify Console. In the Input tab, find Separate datasets and replace the example IDs with these:

Input field Sample data Dataset ID
Dataset IDs from shop 1 19 Headphones.com listings hDS89Nwie71TfmGRz
Dataset IDs from shop 2 98 Audio46 listings n4shucMvwVjRdEKz7

Leave Pair dataset IDs (Pair datasets) empty. The matcher will automatically create product pairs from the two store datasets.

Step 2. Map the correct product fields

The sample datasets store product names under title, while the matcher expects a field called name. Using Attribute mapping here will resolve this issue.

Paste the JSON below into Input attributes mapping:

{
  "eshop1": {
    "id": "title",
    "name": "title",
    "price": "price/current",
    "long_description": "description"
  },
  "eshop2": {
    "id": "title",
    "name": "title",
    "price": "price/current",
    "long_description": "description"
  }
}

Here, eshop1 represents Headphones.com, and eshop2 represents Audio46. The path, price/current, tells the matcher to read the current value inside the price object. Product titles also serve as IDs because they are unique within each store’s dataset.

To display product names, URLs, prices, and available variant details alongside each matching decision, paste the following JSON into Output attributes mapping:

{
  "eshop1": {
    "name_a": "title",
    "source_a": "source",
    "price_a": "price",
    "variants_a": "variants"
  },
  "eshop2": {
    "name_b": "title",
    "source_b": "source",
    "price_b": "price",
    "variants_b": "variants"
  }
}

Fields ending in _a show Headphones.com’s product details, while those ending in _b show Audio46’s, so you can compare their listings side by side.

Step 3. Run the matcher

Under Precision/recall tradeoff (Model configuration), select precision for more reliable matches. Recall finds more potential matches but also returns more incorrect suggestions.

Click Start and approve the Actor’s Full permissions when prompted.

Once the run shows Succeeded, open Output to see the product pairs and their predictions:

  • predicted_match: 1 means the Actor predicts a match, and0 means it doesn’t.
  • predicted_scores: Higher scores indicate stronger support for a match. Use predicted_match to filter the results.

The test returned 1,862 pairs. Checking the predictions against the reviewed product data gave these results:

Result Pairs
Correctly matched the same model 9
Correctly rejected different products 1,824
Incorrectly matched different products 2
Missed pairs of the same model 26
Unresolved after review 1
Total 1,862

The matcher suggested 11 matches: nine were correct, but two incorrectly paired Ananda Stealth with Edition XS.

Step 4. Compare the matched products’ prices

The output contains all 1,862 product pairs the matcher evaluated. Before comparing prices, narrow it to the nine confirmed matches by filtering out rejected pairs and the two incorrect suggestions:

  1. Click Export (top right of the run page), chooseExcel , and selectDownload .
  2. Open the downloaded file, click any cell in the data, then select Data → Filter to add filter arrows to the column headings.
  3. Click the arrow beside predicted_match , uncheckSelect All , select1 , and clickOK . You should now see the11 pairs the Actor predicted as matches.
  4. To exclude the two incorrect pairs, click the arrow beside name_b , chooseText Filters → Does Not Contain , enterEdition XS , and clickOK .

You should now see nine rows, each pairing a Headphones.com listing with an Audio46 listing for the same model. You can compare what the stores charge by reading the two price columns in that row:

  • price_a/current: Headphones.com’s listed price.
  • price_b/current: Audio46’s listed price.

These values are in cents, so divide each by 100 to get dollars. For example, 12900 becomes $129.00. The table below compares three of the nine confirmed pairs, with prices converted to dollars.

Product Headphones.com listed price Audio46 listed price Audio46 minus Headphones.com
Sundara Closed-Back $129.00 $129.00 $0.00
Ananda Nano $319.00 $319.00 $0.00
EF400 $299.00 $529.00 $230.00

Matching the EF400 listings revealed a $230 difference between the stores’ listed prices. With your own catalog, these comparisons can help you easily spot where competitors charge more or less for the same product and decide which prices to review.

Try it with your own catalog #

From 117 listings across two stores, AI Product Matcher checked 1,862 possible pairs and flagged the closest matches. That gives you a starting point for comparing what your competitors are charging for the same products.

To try AI Product Matcher with your catalog, collect each store’s listings in a separate Apify dataset. Then replace the sample dataset IDs and adjust the mappings if your data uses different field names.

Frequently asked questions #

How much does AI Product Matcher cost?

The run in this tutorial cost $0.021, or about two cents, to compare 1,862 product pairs, excluding scraping costs. AI Product Matcher uses pay-per-usage pricing, so your cost depends on the computing and storage resources each run consumes.

How can I compare larger product catalogs?

Scrape each store into a separate Apify dataset and use the same setup shown in this tutorial. To compare your store with multiple competitors, keep your dataset under Dataset IDs from shop 1 and run a comparison with each competitor’s dataset under Dataset IDs from shop 2.

Can the matcher recognize the same product in different colors?

Yes. The matcher is trained to treat different color variants as the same product. For example, it considers a black phone and a white version of the same model a match.

── more in #ai-products 4 stories · sorted by recency
── more on @apify 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/ai-product-matching-…] indexed:0 read:6min 2026-11-18 · —