How do I structure shopping campaigns?

Structure exists to give you control and readable reporting. Every extra campaign costs you volume in each one, so add a split only when you can name the decision it lets you make.

Write the questions down first. Do branded searches perform differently from generic ones. Are my high margin lines getting enough budget. Is the new range working. Is one country carrying the account.

Each genuine question justifies one split. Everything else is complication.

For a store starting out that usually means two or three campaigns, not fifteen. As spend grows, more splits become affordable, because each campaign still has enough volume to be readable.

The opposite failure is just as common. A single campaign containing everything is simple, but it cannot answer any of the questions above, and its average hides which products are earning and which are being subsidised.

The test for any proposed split is this: if the report showed a difference between these two groups, would you do something about it. If not, do not split. Structure that nobody acts on is maintenance with no return.

This is the one split almost every account benefits from, because the two kinds of traffic behave completely differently.

Someone searching for your brand already knows you. They convert at a high rate, cost little, and would in many cases have found you anyway. Someone searching for a product type does not know you, converts at a lower rate, costs more, and represents actual growth.

Blend them and the account average looks healthy while telling you nothing. Worse, the automation will happily concentrate spend where the predicted return is highest, which is your own brand, and you end up paying to reach people who were already coming.

Separate them and each becomes readable. You can set a strict target on brand traffic and a realistic one on generic traffic, and you can see whether new customer acquisition is actually working.

On Shopping the separation is done with negative keywords and campaign priority rather than with keyword targeting, since you do not choose keywords directly. Put brand terms as negatives in the generic campaign, and keep the brand campaign narrow.

Then judge the brand campaign honestly. If it is expensive and the same people would arrive without it, that is a candidate for an incrementality test.

The most common structural fault is products being eligible in more than one campaign without a rule deciding which wins. The campaigns then compete with each other, the reporting splits, and neither number can be trusted.

Fix it with two tools. Inventory filters define which products a campaign may show, so build them from fields in your feed that you control: product type, brand, a custom label for margin band, a custom label for season. Custom labels are worth setting up properly, because they let you slice the catalogue commercially rather than by whatever categories the store happens to use.

Campaign priority decides which campaign enters the auction when a product is genuinely eligible in more than one. It is useful for deliberate tiering, for example a low priority campaign holding everything and a higher priority campaign holding the products you want treated differently. It is only useful when you set it on purpose.

Write down the rule in a sentence per campaign: this campaign covers these products, at this priority, with these negatives. If you cannot state it in one sentence, the structure is too complicated.

Then check it. Look at a sample of products and confirm each shows from the campaign you expect.

Most accounts now run some mix of automated campaign types and standard Shopping, and the mix needs a rule or they will overlap.

The practical question is which campaign type gets which products. Running the same products in both means one will generally take the traffic, and you will not be able to say which approach performed better because the comparison was never clean.

A workable approach is to give each campaign type a distinct product set for a defined period, so the comparison is real. Another is to use the automated type for the broad catalogue and keep standard campaigns for the products where you want direct control over what shows and for which searches.

Whatever you choose, avoid the situation where an automated campaign quietly absorbs your branded traffic and reports an excellent return on it. Use the available brand exclusion settings, and check the search terms data to see what it is actually buying.

And change one thing at a time. Restructuring the campaign types and the bidding and the budgets together produces a month where nothing is comparable to anything, which is a month wasted.

Name campaigns so someone who has never seen the account can read the structure. Country, campaign type, product group, purpose. Consistent order, lower case, no personal shorthand. Six months later that saves an hour every time anyone opens the account.

Put budgets at the level where you make decisions. If you decide budget by product group, budgets belong on campaigns split by product group. If you decide by country, split by country. A structure whose budgets do not match how you actually allocate money forces manual work every month.

Set a monthly routine and keep it short. Check spend by campaign against intention. Read the search terms report and update negatives. Check product level spend for items taking money without earning. Check for new Merchant Center disapprovals. Confirm nothing is out of stock and still advertised.

Then change at most one significant thing, and record the date.

Resist restructuring for its own sake. A modest structure managed consistently outperforms an elegant one that is rebuilt every quarter, because every rebuild resets the learning and erases the history you would have used to judge it.