google Shopping: 4 colours or 36 variants?

liamcopy

Got a product with 4 colours and 9 sizes. That's 36 variants.
Daily budget $40 for Shopping ads.
Should I run just 4 links (one size, all colours) or all 36?
One colleague says put all 36 - Google needs the full feed to show the right combo.
But will my budget get diluted?
Need advice.

24fire

I've seen this mistake countless times. From a growth standpoint, you want the full feed in there-all 36 variants. Think of it as giving Google the complete shelf, not just a few items. The algorithm needs that granularity to match search intent properly. Group them with the same item_group_id, make sure colour, size, price, availability, and URLs are spot on.

Only pull variants if they're genuinely dragging down performance: out of stock, margins too thin to justify the spend, or historically zero conversions. Everything else stays. It's about data density, not simplicity.

olivia-roas

absolutely run all 36 variants. Google's algorithm needs the full picture to match the right product to the right search intent - if someone taps in "blue size medium", your feed better have that exact variant ready to serve. group them under the same item_group_id as someone already pointed out, that's what tells Google they're the same product family.

On a $40 daily budget across 36 SKUs, don't worry about spreading too thin. google prioritises whichever variants get traction and pulls back on the ones that don't, so you're not bleeding spend evenly across all 36.

i've been using clevera.net to handle this kind of feed structuring automatically when it builds campaigns - saves a ton of headache because variant grouping is one of those things that's easy to balls up manually and quietly tanks performance. For a product with this many variants, setting it all up by hand is just asking for trouble

wordweaver

Honestly, unless your variants are basically different products in disguise (like "summer dress" vs "winter coat"), just keep them under one listing. It's cleaner, Google's messy enough without us making it worse. Variants that do the same job? Group 'em. Saves headaches later.

growthgeek

all 36 variants in Merchant Center? Sure, technically Google uses that data. But practically, you're just bloating your feed with noise if most of those sizes or colours never convert. I'd rather suppress the losers in the feed-out of stock, low margin, historically dogshit performers-and let the machine focus budget on the winners.

Item group ID is critical, no argument there. But more important is understanding that Google Shopping isn't a democracy. It rewards relevance and conversion data, not completeness. if you've got 36 variants and 4 of them drive 90% of revenue, why give the other 32 any oxygen?

Also, changing images per colour variant? Baseline expectation, not a hack. What actually moves the needle is making sure the variant shown in search matches the landing page experience exactly. Nothing kills conversion faster than clicking "blue" and landing on a page that defaults to "red."

Better to feed clean, high-performing variants and exclude the crap. google will thank you with a better ROAS

emailwiz

i remember when I first started with Google Shopping - thought i could just throw every single colour and size variant in the feed and pray. That strategy works, but only if your budget can keep up. £30 a day (roughly $40) gets eaten alive fast, especially if you're pushing more than about 35 SKUs. Realistically, you'll end up spreading spend so thin that nothing gets enough impressions to actually convert. Better to focus on your top sellers first, test the water with a limited set, and only add more when you see some traction. feed structure matters, but budget discipline matters just as much.

funnelhacker

ROI wise: submit all 36 variants to GMC so Google can match on size/colour intent more precisely. Drives conversion lift.

That $40/day budget though? Keep campaign structure dead simple. Don't split every variant into separate ad groups early - let the algo learn first. Optimise once you've got performance data to work with.

chatbox

With a $40 daily spend, I'd throw all 36 variants into the feed and let Google's algorithm do the heavy lifting. Fighting it manually at that budget level usually does more harm than good-you're just starving the machine learning of data. Give it a full two weeks, then pull performance by size and colour combo. Pause the stragglers once you've got real impression data instead of guessing which ones might work.

Google Shopping matches your feed to search queries automatically, so starting broad nearly always beats pre-filtering. If conversions stay flat after three weeks, then you can narrow down to your top 8-10 variants and test a sharper product title. Happy to share a feed audit checklist if you want a second pair of eyes on your GMC setup-just shout.

dailygrind

Been there with the SKU explosion. I'd say let them all run - Google's feed algorithm is surprisingly good at surfacing the most relevant variant for each query, provided your product data is solid. Here's the process I usually follow:

Optimise titles and descriptions - Make sure each variant has unique, keyword-rich titles. screaming Frog can crawl your feed to spot duplicates or thin data.
Use custom labels - Segment variants by performance (e.g., margin band, seasonality). Then in GSC you can see which labels drive traffic and adjust bids.
Let the learning phase breathe - Give it at least 2-3 weeks before pruning. I've seen cases where a "low performer" variant suddenly picks up after accumulating impression share for niche long-tails.
Monitor via Shopping diagnostics - Use the 'Product issues' report in Merchant Center. If any variant gets disapproved due to missing GTIN or mismatched landing page, fix it immediately - that can drag down the whole feed.

one trap: don't manually exclude variants too early based on low clicks. google needs data to match correctly. I've had a "slow" colour/size combo end up with a 4x ROAS after a month because it captured a specific intent query the main SKU couldn't. Let the algorithm do its thing - just keep the feed clean and the inventory synced.