A Shopping or product-ads feed is only as trustworthy as the extra fields you attach to each item. Titles and prices usually already live in the store database. The attributes that get added later — a grouping field, a brand override, a flag that an item should stay out of ads — often end up in a Merchant Center rule or a spreadsheet. That works for a week, then prices change, new products arrive, and the sheet is already wrong.
A hardware catalog makes this obvious. A tube of silicone sealant and a multi-part pipe fitting set are very different products, and the extra feed fields you attach to them will not age at the same speed. If those fields live outside the store, they drift.
Put the field on the product
Merchant Center can set extra attributes through feed rules or supplemental sheets. Many shops start there because it is fast. The cost shows up later, when nobody remembers to update the sheet.
A more durable pattern is a column (or a small related table) on the product itself. The feed endpoint just reads it:
const items = products.map((product) => ({
id: product.sku,
title: product.title,
price: `${product.price} ${product.currency}`,
// extra feed attributes come from the store, not from a sheet
feed_attribute: product.feedAttribute,
}));
The idea is simple: if a value changes when someone in the business changes their mind, it belongs in the database, not in code and not in a third-party UI.
Defaults and overrides
Two layers are usually enough:
- A computed default. A scheduled job can fill the field from data you already have, such as brand, category, or stock status. Keep any mappings in settings, not as constants in the repo, so a change is an update, not a deploy.
- A manual override. Some products need a human decision: a new item with no history, a product whose data does not map cleanly to how ads show it, or an item that should stay out of the feed for now.
When you change a lot of rows at once, copy the old values first and keep the rollback next to the update:
CREATE TABLE backup_product_feed_attr AS
SELECT id, feed_attribute
FROM products
WHERE category_id = :category_id;
UPDATE products
SET feed_attribute = :new_value
WHERE category_id = :category_id;
-- rollback:
-- UPDATE products p
-- SET feed_attribute = b.feed_attribute
-- FROM backup_product_feed_attr b
-- WHERE b.id = p.id;
It is a two-minute habit, and a bulk update on live product data feels a lot less scary.
After you change the feed, wait
Merchant Center does not pick up feed changes instantly. If you change a batch of attributes and immediately assume the ads side has caught up, you get a day of confusing reports.
Before you rely on the new values:
- In Merchant Center, filter by the attribute and compare the counts to a
GROUP BYon your side. - Spot-check a few SKUs individually.
- Prefer a quiet hour if the change will affect what can appear in ads.
A missing or misspelled value is easier to find if those products simply do not get the attribute, rather than silently inheriting something leftover from a sheet.
Things that help from day one
- Put the attribute in the database immediately. The spreadsheet phase is usually wasted work.
- Log every change with a reason. Weeks later, "why does this product have this value?" is a question you will want answered.
- One source of truth. If the same field is also maintained as a rule inside the ads platform, the two will disagree.
None of this is advanced ads work. It is plumbing: the feed should read store data, and you should give Merchant Center time to catch up. If you already keep these fields somewhere else, a first step is just listing where each attribute currently lives — database, sheet, or a platform rule — and picking one place.
Top comments (0)