Advertise Only the Cheapest Variant

Stop paying to advertise eight near-identical rows of the same product. One Group products rule keeps the cheapest product per group, and one ID mapping keeps its identity stable across imports.

Updated November 6, 2026

If your feed carries several rows for what a shopper sees as one product, such as every size of a shirt or every offer for one barcode, you're spending budget saying the same thing several times over. This guide keeps one product per group, the cheapest one, and keeps its identity stable between imports.

What you'll end up with#

A catalog where each product family appears once, represented by its lowest-priced member. If prices move and a different member becomes the cheapest, the catalog keeps the same product and updates its price, so the ad network sees no new product and no learning is lost.

1. Find the value that identifies a product family#

You need one value that's the same for every row in a family and different between families. Good candidates are:

  • a variant family id, often called item_group_id, parent_sku, style_code or model
  • a barcode (GTIN/EAN), when your rows are competing offers for the same physical item
  • a combination like brand + style code, when neither alone is unique

Open the catalog from your project's Catalogs list, go to the Rules tab, and check the column list, or the Mapping tab's column rail, to see what your feed actually has.

If nothing identifies a family, stop here: without a shared value there's no group to reduce, and you'd be picking one product out of your whole catalog.

2. Add the Group products rule#

  1. On the Rules tab choose + Add rule, then set the action to Group products. The grouping editor opens.
  2. At the top of the dialog, choose Keep one product per group.
  3. Under Group by, enter the family value you picked in step 1. Leave the matching mode on Smart unless you're grouping on barcodes, where Barcode is the better choice.
  4. Under Which product to keep, choose lowest and enter your price column.
  5. Leave the computed fields empty. You don't need any for this job.
  6. Check the live count at the bottom of the dialog. "~50 products → ~12 kept in a sample" means it's working; "~50 products → ~50 kept" means your group key is unique per product and nothing is grouping.

Give the rule a name while you're there. "Cheapest variant only" reads better than "Unnamed rule" the next time you open the tab.

3. Point the ID at the group key#

Before you leave the dialog, look under Which product to keep. The editor warns you there when the kept product's ID doesn't come from the group key. Take it seriously.

If your ID is mapped to each row's own id, the catalog id changes the moment a different variant becomes the cheapest. Every ad network reads a changed id as a brand-new product: fresh learning, lost history, and a spike of "new" products in your reports.

Choose Use group key as ID in the warning, confirm, then choose Done. Renable maps the ID to your group key, with all of its parts if you grouped on more than one, and replaces any fallbacks the ID had. You can also do this by hand on the Mapping tab. Every member of a group now produces the same id, so whichever one is kept, the id stays put.

If there's no warning, choose Done.

Changing the mapping changes your products' ids once, when you publish, so that one time destinations see them as new.

4. Publish, then check#

Publish the catalog. The publish confirmation estimates how many products will be grouped into other products. If it says nearly your whole catalog, your group key is too coarse.

After the import runs, open the Activity tab and expand the run:

  • Products seen should be roughly your old count divided by the average family size.
  • Grouped into other products should carry the difference.
  • Nothing should show up under Excluded because of this rule, since grouped products are counted separately.

To check one family, use Trace on the Rules tab. Pick a product that was grouped away, and the trace confirms it was grouped rather than excluded and, when it can identify it, names the product kept in its place.

If you change your mind#

Delete the rule, or switch its enabled toggle off, map ID back to each row's own id on the Mapping tab, and publish. Every product comes back on the next import. If the ID still comes from the group key, a whole family shares one id and only its first row is imported. Edits you made to products that went away are kept for 30 days.

Variations on the same rule#

  • Keep the most expensive one instead: same rule, choose highest.
  • Keep the one that's in stock: point the selector at a stock-count column and choose highest.
  • De-duplicate identical rows: group on the barcode with Barcode matching and keep the lowest price.
  • Keep every product but show the family's lowest price: you don't want this guide. Use Add computed fields to every product with a Smallest of price. See Grouping products.
  • Your family value is blank on some products: with the ID mapped to that value they have no ID, and the import leaves them out. In the same rule, condition on the family column being blank, Set field it to the product's id in the main branch, and move the Group products action into the "Otherwise…" branch. Products with a real family id are reduced to the cheapest; the rest keep an id of their own and pass through untouched. Grouping products walks through the shape.

Or ask Aimée#

Try "only advertise the cheapest variant of each shoe, and keep the id stable". Aimée can add the rule and the ID mapping, then publish once you approve.

Related articles#