Google Ads for e-commerce: campaign management and optimisation

What decides whether a Google Ads account makes or loses money for an online store is not the bidding settings. It is how the account is divided, whether anyone reads the search terms, and the quality of the product data feeding the shopping campaign. This guide takes those three in order, and says at each step when to intervene and when to leave it alone.

1. Account structure: what has to be separated from what

There is one rule in the structure argument: do not keep two things you cannot compare inside the same campaign. The average that comes out does not reward the half that works, it hides the half that does not.

• Brand apart from non-brand. Brand searches are cheap and convert well. Kept in the same campaign, they make the whole account look better than it is.
• Split by margin band. Optimising an 8 percent margin appliance and a 60 percent margin accessory toward the same target is meaningless.
• New customers apart from existing ones. Remarketing always returns more, and mixed into the same account it eats the budget of the campaigns meant to grow you.
• Do not overdo the campaign count. Automated bidding learns separately in every campaign. Splitting the budget twenty ways means none of them ever learns.

2. The search terms report: the most useful screen in the account

A keyword is what you meant to buy. A search term is what you actually bought. The gap between them widened as broad match and automated campaigns spread, and reading this report weekly is still the cheapest improvement available in any account.

• A negative list is not written once and left. “free”, “second hand”, “how to make”, “jobs” and out-of-category brand names accumulate again every week.
• Give converting long-tail terms their own group. Low-volume, high-intent terms perform noticeably better once they have ad copy written for them.
• Check for overlap between campaigns. When the same term matches in two campaigns, the learning data for both is split.

3. A shopping campaign is really your feed’s performance

You do not write ad copy in a shopping campaign. Google takes your product title, image, price and attributes straight from the feed. So the practical meaning of “optimising the shopping campaign” is fixing the product data.

• Title: the most heavily weighted field in matching. Brand, model, distinguishing feature, colour, size works in most categories. “Stylish Shoe” matches no query at all.
• Category: Google’s taxonomy runs to more than 5,500 branches. The wrong branch drops the product into the wrong competition.
• Identifiers: with gtin, brand or mpn missing, the product disappears from several comparison surfaces entirely.
• Disapprovals: every rejected item in Merchant Center is a product that is not in the campaign. That list is a weekly read, not a monthly one.

To measure where your feed stands on those four, use the free feed analysis: more than 30 checks run per product, the catalog is graded A to F, and the gaps costing you the most are ranked. The source feed is never touched.

4. Performance Max: what you gain and what you stop seeing

Performance Max reaches the whole inventory from one campaign and usually performs well. In exchange you give up some visibility, so what makes it controllable is not its settings but what you feed into it.

• Exclude brand search. Otherwise the campaign reports sales that were coming anyway as its own achievement.
• Separate the product groups. With the whole catalog in one asset group, budget flows to a handful of bestsellers and the rest is never seen.
• Supply the image and text assets yourself. Left to automatically generated assets, the campaign can pick the most incidental image on your site.
• Use the new customer goal. It is the only way to tell the campaign to reward a new customer rather than an existing one.

5. Bidding strategy and the learning period

Smart bidding is only as good as the conversion data behind it. A strategy running on too little data is guessing, and it tests that guess with your budget.

• Do not set a target while conversion volume is low. Collect data first with a maximise-conversions strategy, then add the target.
• Move targets in steps. A large jump pushes the campaign back into learning and costs you a fortnight.
• Plan for the season. Peak weeks do not behave like last month. Set targets from the same weeks last year.
• Do not intervene daily. A team touching the account every day resets the algorithm’s learning continuously.

6. Ad copy and assets

In responsive search ads Google assembles the headlines itself, and it can only do that meaningfully if the headlines you supplied are genuinely different from each other.

• Give it distinct angles: the benefit, the price position, delivery, warranty, range. Five rewrites of one sentence is not a test.
• Point sitelinks at real pages: a category, a promotion, the delivery terms. Four links to the home page take up space and do no work.
• Spend the callouts on what a shopper is already checking: free delivery, returns window, same-day dispatch.
• Test the offer, not the wording. The difference between “10% off” and “15% off your first order” teaches you something.

7. Is the measurement right?

This is the question to answer before any optimisation discussion, because in a badly measured account every improvement is accidental.

• How far apart are the platform and the ledger? Anything over about 10 percent needs an explanation. The usual cause is the same purchase being sent from two sources.
• Is the conversion value real? In an account sending a fixed value, smart bidding rewards your lowest-margin products.
• Are returns deducted? In high-return categories, optimising on gross revenue systematically grows the wrong products.

8. The weekly routine

A Google Ads account is not set up and left, and it is also not poked at daily. One hour a week, in this order:

• Merchant Center disapprovals, because a failure here costs you inventory rather than budget.
• The search terms report and this week’s new negatives.
• Return by campaign, reading brand and non-brand separately.
• Campaigns hitting their budget cap, and campaigns unable to spend theirs.

Frequently asked questions

How long does Google Ads take to work for e-commerce?

A brand campaign reads within days. Shopping and non-brand search usually need a few weeks of steady data before automated bidding has learned anything. Large setting changes during that period restart the process, and that is why most accounts learn late.

My shopping campaign is live but getting no impressions. Why?

Check three places in order: are the products approved in Merchant Center, is the right product group selected in the campaign, and is the bid below what the category costs. In our experience the first is the most common, and it usually traces back to one missing required field.

Should I really pay for my own brand searches?

If competitors are bidding on your brand, yes, because otherwise you hand them traffic that was looking for you. If they are not, you can cut the budget and let the organic result do the work. That decision is made by looking at the search result itself, not at the campaign dashboard.

If you are not sure which part of your account is a data problem and which part is a campaign problem, our advertising side makes exactly that distinction. For a quick look, write to us.