Google documents what gets your product in. What makes it win, you test.
I took each of my eight feed rules back to Google’s own documentation, and looked for anyone’s published test of a feed change. Eligibility is documented: miss a required field and the product is out. Performance is not. Google’s own claims about what feed data does come without a method, so I test them.
Sources checkedOctober 2026
Last revised12 October 2026
Next reviewJanuary 2027
ByChuck Aikens
one-person consultancy for consumer product brands
15 years
Fill every gap Google documents. Then test what you change.
The rules
All 8 rules. Three shown in full.
1RULE 1 OF 8 · REQUIRED
Fill the required fields first
Every product carries every field Google marks Required, plus the conditional fields that apply to it, plus shipping. Nothing optional gets attention until these pass.
What backs it
Google grades every attribute Required, It Depends or Optional. Of Required it says: “Submit this attribute. If you don’t, your product won’t be able to serve in ads and free listings.”Google Merchant Center Help, Product data specification
That is not a ranking penalty. It’s a closed door, and it’s the one feed problem with a documented, certain cost, so it comes first. The framing is my reading of the specification.
7 steps to apply this are in the full Guideline.
2Get identifiers rightIn the full Guideline
3Own your category treeIn the full Guideline
4Tie variants together and fill the apparel fieldsIn the full Guideline
5RULE 5 OF 8 · REQUIRED | EXPERT PRACTICE
Images and copy that match the page
The main image is clean, the title leads with what matters, the description does the same, and nothing in the copy is promotional or absent from the landing page.
What backs it
Images: Google requires “at least 500 x 500 pixels for all product images,” with enforcement beginning January 31, 2027, and says “Don’t include promotional text, watermarks, or borders.”Google Merchant Center Help, Product data specification
Titles: “Use all 150 characters.” “Put the most important details first.” Google adds that users “will usually notice only the first 70 or fewer characters of your title, depending on screen size,” and says not to include promotional text.Google Merchant Center Help, Title and structured title
The rest is mine. 1500 pixels is the image size I aim for, and 4 to 6 is the number of highlights I use: enough to make the case, few enough that each one is read. Both are working rules, not Google’s.
5 steps to apply this are in the full Guideline.
6Fix at the source, patch with a supplemental feedIn the full Guideline
7Keep the feed and the site in syncIn the full Guideline
8RULE 8 OF 8 · EXPERT PRACTICE
Test before you trust
Any change beyond the required fields gets tested with a randomized item split before it rolls out. One change at a time, read on paid metrics, run outside the holiday peak.
What backs it
Google, Perplexity and OpenAI all say richer feed data helps; none publishes how it measured that, and I found no public controlled study of a Google Shopping feed lever. The 20 percent above is the pattern.
A before-and-after comparison can’t separate your change from seasonality or a competitor’s sale. Randomizing items can. FeedX, an open-source method on Google Marketing Solutions’ GitHub, is built on that distinction. Its README says its crossover design averages out seasonality “assuming the washout periods are set correctly,” and that “This is not an official Google product.” It is a method, not a result: I found no published result from it that I could check.FeedX README · Google for Developers blog, May 2025
FeedX also uses CUPED, a noise-reduction step from Deng et al. (Microsoft, WSDM 2013). On a few Bing engagement metrics, such as clicks and visits per user, it cut variance by about 50%; on revenue per user it cut under 5%. Bing split users, not products, so 50% is a sense of scale, not a forecast for a feed.Improving the Sensitivity of Online Controlled Experiments by Utilizing Pre-Experiment Data
8 steps to apply this are in the full Guideline.
The short version
Fill the gaps, then test.
Documented rules decide whether a product serves at all. Past that, a claim about performance stays a hypothesis until a randomized test says otherwise. What each rule asks, and what I’ll claim it does, is set out rule by rule.
Measured: a published test, population and date stated.
Platform says: stated, no method shown.
Expert practice: my judgment, my reason.
Unproven: repeated, nothing I could check.
Open: sources conflict, or never tested.
The grade is the honest part. The rule is the point.
What I’d tell you in a meeting.
Clients get the full Guideline
How to do it: 34 steps across the eight rules
Measuring honestly
Where to start
The audit checklist
What clients get
Three rules and one myth of eight. Clients read the whole Google Feeds Guideline on app.tymoo.ai, updated quarterly. When Google’s rules or the research move, I update the rules and say when.