Google Ads for eCommerce: FAQ

Product listings with an image, price and store name that appear on Google Search, the Shopping tab and across Google's network. They're generated from your product feed in Merchant Center, not from keywords you write. That means your feed data, not your ad copy, decides where you show up.

For most product businesses, yes, because the intent is already there: someone typing a product name is closer to buying than someone scrolling a social feed. Whether it's profitable depends on your margin and your breakeven ROAS, not on the channel itself. If your gross margin is under 25% and your average order value is low, the maths gets tight fast.

You pay per click, and clicks in eCommerce categories typically land between $0.30 and $2.50 depending on competition and country. The more useful number is cost per acquisition against your margin. I work with brands spending $20k to $250k a month, because below that there usually isn't enough data to make structural decisions with any confidence.

Most healthy eCommerce accounts run all three, with clear roles. Shopping and PMax carry product demand. Search carries category and problem-level queries plus brand defence. The mistake is letting PMax absorb everything, because then you can't tell which part is working.

Create a Merchant Center account, verify and claim your domain, submit a product feed, pass the review, then link Merchant Center to Google Ads and build the campaign. The order matters: campaigns built before the feed is clean will spend on products Google doesn't properly understand.

Both have native apps or plugins that push a feed automatically. Use them as the base, then add a supplemental feed for the fields the app fills badly, usually titles, product types and custom labels. The default export is a starting point, not a finished feed.

The frequent causes are mismatched price or availability between feed and landing page, missing GTIN or brand, missing shipping and returns policy, image issues, and 'misrepresentation' triggered by thin policy pages or unverifiable claims. Fix the underlying data, not just the flagged item, or the same disapproval returns in the next crawl.

They're the strongest lever you have. Shopping matches queries against feed text, so a title structured as brand plus product plus key attribute plus size or colour will pick up queries a generic title never sees. Rewriting titles across a catalogue routinely moves impression volume more than any bidding change.

Not positively. You can only exclude with negative keywords and shape matching through the feed and campaign structure. That's why feed work and search-term exclusions are the real 'keyword strategy' in Shopping.

In this order: fix feed data, exclude products that spend without returning, separate brand from non-brand so blended numbers stop hiding the truth, then adjust bidding against real breakeven. Bidding is last because it's the fastest to change and the least likely to be the actual problem.

Split by intent and by margin, not by whatever the catalogue tree looks like. Brand separated from non-brand. Best sellers separated from the long tail so the top 20% of SKUs aren't subsidising products that never convert. Testing budget kept in its own line so it can't quietly eat the scaling budget.

Across Shopping, Search, YouTube, Display, Discover and Gmail, and the report doesn't cleanly show you the split. Contain it: brand exclusions, account-level negatives, tight asset groups and feed-only tests where the channel mix needs proving. Once it earns trust, loosen it.

Very often, yes. Without brand exclusions PMax will buy the cheapest conversions it can find, which are the ones you'd get anyway. It looks like great ROAS in the dashboard and flat revenue in the bank account. Excluding brand terms is the first check I run on any PMax-heavy account.

Feed and tracking fixes show up within days. Structural changes need two to four weeks to re-stabilise learning. Anything judged on a three-day window is noise, not a result.

There's no universal number. Good means above your breakeven, which is 1 divided by your gross margin. At 40% margin your breakeven is 2.5×, so a 3× ROAS is profitable. At 20% margin your breakeven is 5×, and that same 3× is losing money on every order. Run your own number in the breakeven ROAS calculator before you set any target.

Start from breakeven, then set the target above it by the margin you want to keep, and account for repeat purchase if your data supports it. Setting Target ROAS from whatever Google suggests is how accounts end up optimising for a number nobody in the business chose.

Usually a mix of three things: brand traffic inflating the blended number, conversions double-counted between GA4 and Google Ads, and revenue booked before returns and discounts. The dashboard is measuring attributed revenue. Your accountant is measuring money. When they disagree, the dashboard is the one that's wrong.

Enough for the account to produce statistically meaningful data in a reasonable window, which in most eCommerce categories means at least 30 to 50 conversions a month per campaign that you want to optimise separately. Below that, scaling decisions are guesses dressed up as strategy.

Smart Bidding, in almost every case, but only after conversion tracking is trustworthy. Automated bidding on broken tracking optimises faithfully towards the wrong outcome, and it does it faster than you can catch it.

Different attribution models and different windows. Google Ads credits the conversion to the click date and uses its own attribution; GA4 reports on the conversion date with its own model. Some difference is normal. A gap above roughly 20% usually means a tagging problem, not a modelling one.

Audit what's firing: one purchase conversion counted as primary, deduplicated, with dynamic value passed from the order, no thank-you page refresh double-count, and no secondary actions like 'add to cart' quietly counted as conversions. Then verify against real orders in your store backend, not against the tag assistant.

It helps most where consent banners and browser restrictions eat a meaningful share of your data, which in the EU is most stores. It isn't a fix for broken tracking logic. Get the client-side implementation correct first, then move it server-side for durability.

With consent mode v2, denied consent means modelled rather than observed conversions. Your reported numbers stay usable but become estimates. The important part is that it's implemented correctly, otherwise you lose the modelling as well as the observed data.

Spend, non-brand ROAS, cost per acquisition against target, and what changed in the account plus why. Three sentences a week beats a forty-page deck nobody opens. That's the Friday decision log I send to every client.

DTC eCommerce brands spending roughly $20k to $250k a month on Google Ads. Six clients maximum at any time. No outsourcing, no account managers relaying between us.

The Decision Audit is free, 30 minutes, no pitch at the end. The Written Account Diagnosis is $1,500 one-time. Ongoing management starts at $2,500 a month. Final scope is set on the call.

Yes. I often step in as a consultant during the notice period, so the handoff isn't a cold start once the old agency is gone.

Book a Decision Audit. Thirty minutes, I look at the account live and tell you what I see. If it isn't a fit, I'll say so on the call.

Yes, two ways. Referral: you send me a client, I work with them directly under my own name, you earn a monthly fee for as long as they stay. Whitelabel: I run the account under your agency's brand, your reports, your client relationship, on an agreed margin. Details on the partnerships page.