In Shopping there are no keywords.
Your product data is the campaign.

Because in Shopping you do not choose the queries you appear for. The platform matches the shopper's query against your product data, so titles, categories, identifiers and images decide what you are eligible for, and the bid only decides whether you compete once eligible. The product data specification sets out every requirement. This is why we fix the feed before spending, and why feed work usually outperforms campaign work in inherited accounts. The Shopping specifics are on our Shopping page.

It treats every revenue dollar as identical when your margins are not. A campaign optimized on return on ad spend grows whichever products convert most readily, and in most catalogs those are the thinnest-margin lines — so the metric improves while contribution falls. We ask for margin by product line before structuring anything, and where it is not available we say plainly that we are optimizing a proxy until it is. The catalog-wide version of this argument is on our ecommerce program page, and the conversion-value mechanics that make it possible are documented by Google.

Yes, once the signal underneath it is verified. Google's Smart Bidding documentation is explicit that these strategies optimize toward the conversions an account reports — so they are extremely effective at pursuing whatever you told them to value. Enabling them over unaudited conversions scales an existing error efficiently. Verify what is counted, correct the values, then automate. That sequencing is what our AI-first approach means in a paid account rather than a claim on a slide.

Usually a mismatch between feed and page, or a missing required attribute. The feed requirements and shopping policies set out both. What most teams miss is that items can be quietly suppressed rather than formally disapproved, producing no alert at all — which is why we compare the feed against the live page rather than only against validation. The catalog reconciliation runs through our ecommerce practice.

Because they measure different things. Ad platforms report conversions they believe they caused using their own attribution windows; your commerce platform records orders that happened. Google's conversion tracking documentation explains what it records and what it does not. The gap is normal — what matters is that somebody reconciles monthly and reports the difference rather than quoting whichever number flatters the work. We also subtract returns, through your own order data.

The way customers describe the product, with the attributes they actually search on carried early. An exported title typically reads as an internal SKU string — brand, model code, a reference number — which competes for queries nobody types. Rewriting it changes which searches the product is eligible for, so it is a copywriting task with the impact of a bidding strategy. The category guidance matters alongside it, and the same language work improves organic performance through ecommerce SEO.

By contribution rather than by catalog category, which is the more common default. Grouping products by how much you keep on them lets budget follow profit and lets you cap thin-margin lines deliberately instead of starving them by accident. Google's account structure guidance covers the mechanics; the segmentation decision is a business one we work out with you first. The same principle applies wherever we manage media, including paid search.

In some categories they change it entirely. Return rates vary enormously by product type, and a campaign that looks strong on gross revenue can be unprofitable once returns are subtracted — particularly in apparel and anything sized or fitted. Because returns arrive weeks after the order, they never appear in the ad platform's view at all. We track them per product group and net them out before calling anything a result, using the order data in your own systems and the same reconciliation standard as remarketing, against what conversion tracking can and cannot see.

In two places. Inside the account, automated bidding and campaign types now make most of the buying decisions, which makes signal quality decisive rather than merely useful. Outside it, assistants increasingly summarize product options before a shopper reaches any store, assembling those summaries from the same structured product data that governs Shopping eligibility. So clean feed work pays in the auction and in channels you are not bidding in. Our position on what can and cannot be promised there is on the GEO page, and the underlying markup is documented by Google.

Almost never. Most catalogs contain products that cannot be advertised profitably at any bid — margin too thin, price uncompetitive against the price benchmarks Merchant Center reports, or a category where acquisition costs more than the item earns. Advertising them because they exist is how budgets get consumed by items that were never going to pay. We identify those lines by name and recommend excluding them, which reduces the media we manage. Google's ecommerce guidance covers the organic alternative for products worth having visible but not worth buying traffic for, which is exactly the work our ecommerce SEO practice handles.

Three things you can verify. We fix the feed before spending, even though it delays launch. We ask for margin by product line before structuring anything and tell you plainly when we are using a proxy. And we reconcile against your commerce platform monthly net of returns, including when the honest number is smaller than the dashboard. We are also built as an AI-first agency rather than a traditional shop with AI added on. Start with the A.R.C. Report, or read the FTC's advertising guidance against whatever your current agency claims.