In Shopping you do not choose the searches: the feed chooses them. That makes the search terms report and negative keywords your only direct control over where you show. And it adds a reading that does not exist in text ads: the Shopping search terms report is a diagnosis of the feed — it tells you what vocabulary Google is understanding from your titles.
Reading the search terms report in Shopping
The same report as for text (Campaigns → Insights and reports → Search terms), but with no keyword column: each term appears with the product (or group) it triggered. Three readings:
- Irrelevant terms triggering a product: the title has an ambiguous word («desk» triggering «office desk» for a mixing desk), or a confusing attribute. Negative… and fix the title.
- Expensive generic terms («running shoes») with poor conversions: they belong in the high-priority campaign with a demanding ROAS (lesson 2), or in negatives if they never convert.
- Competitor terms or brands you don't sell: straight to negatives (unless you have a conquesting strategy).
And a fourth, the most valuable: terms that convert and do not appear in your titles. That is vocabulary the audience uses and your feed does not: it goes into the titles (module 6), not into a keyword list.
Negatives by priority tier
The three-tier architecture is maintained with negatives:
| Campaign | Negative list |
|---|---|
| Generic (high) | Model and brand names of the products + your brand + universals |
| Product (medium) | Your brand + universals |
| Brand (low) | Universals only |
The "models and brands" list is built from the feed (the brand and
mpn/model attributes) and is maintained as catalogue is added. The
"your brand" list includes variants and the usual misspellings.
The universal list for an ecommerce business
On top of the Basic ones (free, jobs, course…), in Shopping it is worth negativising from day one:
- Second-hand and rental: used, second hand, pre-owned, rental, hire, gumtree, vinted, ebay (if you don't sell there).
- Information and support: manual, instructions, spec sheet, warranty for (when it is a query), repair, spare part (if you don't sell spares), compatible with (case by case).
- Pure comparison: vs, versus, difference, reviews, review, test (nuance: sometimes they convert; look at the data).
- Free and downloads: free, gratis, pdf, download, apk.
- Channels you are not: amazon, aliexpress, temu (unless you want to fight that intent on price).
- Geography outside your delivery area: countries/regions you don't ship to.
Thirty well-chosen roots prevent thousands of listing clicks that will never buy.
The feed ↔ negatives cycle
In Shopping, every negative has a twin question: why was my product showing for this? If the answer is "the title is ambiguous", the negative is a plaster and the title is the cure. The healthy cycle:
- Irrelevant term spotted → negative (stops today's spend).
- Review the title/attributes of the product that triggered it → fix it (avoids the whole family of terms tomorrow).
- Good term with no vocabulary in the feed → add it to the title.
- Repeat weekly; quarterly, run n-gram mining (module 1) over the Shopping terms: the vocabulary patterns show up on their own.
Warning signs in the report
- Lots of single-word generic terms («shoes», «tap»): structure by priority is missing.
- Terms in another language or from another country: check the feed (language, country of sale) and the targeting.
- Terms with your brand in the generic campaign: the brand negative or the brand campaign is missing.
- Products triggering searches from another category: a title with a polysemous word.
💡 Ninja trick: the Shopping Ninja applies to Shopping terms the same scenarios SQONS applies in text (cost with no conversion, semantic irrelevance with AI) and adds the reading that only exists in Shopping: it links each bad term to the product and the title that triggered it, and proposes the title fix alongside the negative. The plaster and the cure, together.
What you should remember
- The Shopping search terms report is a diagnosis of the feed.
- Negatives by priority tier (models + brand in high; brand in medium; universals in all).
- Universal ecommerce list: second-hand, support, comparison, free, channels, geography.
- Every negative has a twin question: why was it showing? → fix the title.