Last year I was consulting on an account where the agency had set up a bestsellers campaign on Google's advice. They pulled a static list of bestsellers from GA, threw them in a separate campaign and called it done. They were double-bidding on the same inventory and where none the wiser that they had totally butchered the execution.
The campaign never stood a chance.
This seems like a one-off, but it’s not. There’s a system working in the background.
It goes like this: a big agency tests a shopping approach for their largest client - think over a million SKUs. It works, so they package it up and roll it out to every e-commerce client they have. Google sees it, loves it and starts telling every other agency to do it.
Nobody tells you whether it makes sense at a smaller scale, for different product categories, or how to actually execute it. So I will.
Here are 3 common approaches to Shopping Structure - and when to use them.
Split by Category Each category is its own campaign.
Do it if: you need to do a different target ROAS across your product categories (e.g. margins differ, growing category needs extra support etc.) or purchase behaviour is meaningfully different - high vs. low seasonality, long vs. short consideration periods or different target markets.
Don’t do it if: it's just for reporting clarity. You can still build category reports inside a single campaign (or across multiple). Splitting for visibility isn’t worth the hit to efficiency.
Split by Margin Each margin tier is its own campaign.
Do it if: You have high margin variation in your product mix - and this isn't just category differences, it more commonly shows up in markdown product. Google tends to eat up budget with your lowest margin products as strike-through pricing on ads tends to drive a higher CTR. Start with splitting out your full-price product from markdown and if you have real depth in markdown, go further and split new-to-sale and final-sale too.
Don’t do it if: you care more about revenue than profit or if your COGS data isn’t clean.
Split by Performance Products are grouped into performance tiers - best, mid, low and zero impressions - each in their own campaign.
Do it if: your seasonality and product performance is relatively consistent. You have a lot of SKUs and genuinely believe you can do what Google's algorithm is trying to do better.
You have zero impression products you want to force Google to spend on - I like this approach, but make sure you're targeting new product, not old stock that's slow for a reason (e.g. low size options left).
Don’t do it if: your bestsellers mix changes often (sometimes the bestsellers of yesterday are the dogs of tomorrow).
The Verdict
Accounts that split by Performance will still usually end up spending the most on lower performing products - there are just more of them and Google skews towards volume. Accounts that split by Category can end up with more campaigns than is efficient.
Split by Margin is my default because it's the only approach where the logic is airtight: different margins should have different ROAS targets. Full stop.
Final takeaway
Your inventory split needs to make sense for your business and product.
Don’t apply cookie-cutter segmentation solutions without understanding the mechanisms behind it. And always, always, always, review the performance of the new structure to make sure shopping is converting better than it was before (do an on/off/on analysis if your performance fluctuates with frequent product changes).
In saying all of this, if you're not even hitting 50 conversions per campaign in 30 days, don't split anything at all. More structure just means less data and inefficient bidding.
Didn’t rate this article? Hit reply and tell me why.
Cheers,
Sarah Arvela Webb
Brought to you by a bowl-sized mug of English Breakfast tea and a side of agency tea.
