Product feed in Google Shopping: why it explains 80% of the problem

When a Shopping account underperforms, the first thing anyone touches is the bidding strategy — raise the tCPA, switch to target ROAS, try Maximise conversions. It takes weeks, sometimes months, and the problem stays exactly where it was. The cause is almost always the same: the catalogue Google receives does not describe your products the way customers search for them.

There are no keywords in Shopping. What decides which searches you show up in is the text of your feed. The feed is not a technical formality of the integration: it is the targeting. If you first want to see which part of the catalogue can carry spend and which part eats it, the catalogue calculator splits your SKUs into winners, potentials, losers and sleepers using your own numbers.

Three signals that the problem is the feed, not the bid

• Low impressions with the budget left unspent. The platform cannot find enough searches to place your products in. It is not a budget problem: it is a mismatch between what Google reads in your feed and what people type into the search box.
• Low CTR with good placement. The product shows and nobody clicks. The title and the image don't match what the user expected to see: Google surfaced them by semantic proximity, not by real intent.
• ROAS scattered across products at the same price. Two items at £49 performing radically differently. When price, margin and category are equivalent, the difference usually sits in the quality of the title, the attributes and the image, not in the product.

All three read the same way on Meta: the Advantage+ Shopping catalogue runs on the same file, with the same gaps.

The five fields where the money is lost

1. The title copies the internal product name. "Ref. 4821 — Blue M" tells the system nothing. The customer types "women's running t-shirt navy blue". The title is the highest-weighted field in Shopping and order matters: brand and product type first, attributes after, within 150 characters (the first 70 are what actually shows in most formats).

2. The category (google_product_category) is blank or wrong. Without it Google infers, and when it infers it enters the wrong auction. It is a closed field from an official taxonomy: pick the deepest leaf that describes the product, not the generic branch.

3. Variant attributes are not in separate fields. Size, colour, material and gender buried in the description, or simply absent. Without them Shopping filters don't work and variants compete against each other instead of grouping.

4. The image is the studio shot, not the one that wins the click. The main image rules: neutral background, no logos and no overlaid promotional text — Shopping policies reject those without it ever showing up in the campaign report. Context-of-use shots belong in the additional images, which do get shown on the product listing.

5. Identifiers are missing (gtin, mpn, brand). Without a GTIN, Google cannot match your product against the other offers for that same item, and you lose specific-model searches, the highest-intent ones in the funnel.

The report almost nobody opens

Merchant Center has a feed diagnostics section: disapproved products, demoted products and products with quality warnings. In accounts that have been running for a while it is common to find a meaningful share of the catalogue with some kind of issue.

The expensive part is that none of those problems appear in the campaign performance report. A disapproved product generates no impressions, no clicks and no red row: it simply does not exist, while the budget spreads across the ones that do pass the filter. From inside the campaign, a store with 25% of its catalogue disapproved looks exactly like one with the whole catalogue approved — only smaller.

The ten minutes to spend before touching a bid

1. Download the full feed exactly as Google receives it, not the back-office catalogue.
2. Open Merchant Center diagnostics and separate disapproved, demoted and warnings. Disapproved items are potential spend that doesn't exist; demoted ones compete with one hand tied.
3. Take your 20 best and 20 worst performing products and compare their titles field by field. The difference is usually obvious.
4. Count the gaps: how many SKUs with no category, no GTIN, no variant attributes. That percentage is the real ceiling of your account.
5. Only then, touch the bid. With the feed fixed the same strategy performs differently, because the system is finally competing in the right auctions.

This order matters especially in Performance Max, where the feed is not one more asset source: it is the one deciding what the campaign shows across every Google inventory. With a poor feed, PMax has no signal to optimise and drifts to the cheapest traffic it can find.

What the feed does not fix

A flawless feed does not rescue a catalogue with no margin. If the product cannot sustain the minimum profitable ROAS, optimising the feed just makes you sell at a loss faster. The full order is: first which products can pay for advertising, then making Google understand them, and only at the end how much you bid for them.

It doesn't fix measurement either. If orders don't come back to the platform with their real value, the algorithm learns from the wrong number no matter how perfect the catalogue is. And it doesn't fix creative: a weak CTR on a well-built listing is a product and price problem, not a feed one.

Frequently asked questions