Learn Google Ads for Ecommercefrom Operators,Not Course Sellers

Lesson 1

What Changes Once Your Store Already Sells

At six figures a month the beginner advice stops applying. You already have brand demand, real margin numbers, stock limits and a season. So the wins move. They come from account structure, product-level data and honest measurement, not from finding one magic campaign setting. The three levers that move an established ecom account are how you split brand from non-brand, how you group products by margin, and whether your conversion data is complete enough to bid on.

A store doing 10,000 a month and a store doing 300,000 a month have almost nothing in common inside Google Ads. The small store is trying to get a signal at all. The big one already has signal, and its real problem is that the signal is pointed at the wrong products.

We measured 106,400 products across 137 stores running Google Ads. In fashion, the top 1% of products carried 45.2% of all ad revenue and the top 10% carried 85.6%. In home decor the top 10% carried 91.0%. And 71.7% of products that had ad spend sold nothing at all in their 30-day window, eating 14.7% of the budget in fashion and 29.5% in home decor.

So the question for an established brand is never how to make Google Ads work. It is why a chunk of the budget is sitting on products that have never sold, and what stops that happening next month.

Two other things change once you have scale. You have stock, so a product you scale and then sell out of costs you twice: the lost sales, and a campaign that loses its best performer mid-flight. And you know your real gross margin per product, which is the single input Google cannot see unless you hand it over deliberately.

The nine lessons after this one are all versions of the same job: handing Google better information. Better product data, better structure, better conversion signal. Do that and the bidding takes care of itself. Skip it and no bid strategy in the platform saves you.

Established accounts rarely fail on settings. They fail because the money is spread evenly across a catalog where the results are anything but even.

Go deeper: The complete Google Ads strategy for ecommerce >

Lesson 2

How Should an Ecom Brand Structure Its Google Ads Account?

Split brand from non-brand before anything else. Run one Search campaign on your own brand name, keep every other campaign out of it with account-level negatives, and judge growth on the non-brand side only. Brand search converts at a rate nothing else touches, so leaving it mixed in makes a flat account look like it is growing and hides the campaigns that are actually losing money.

People searching your brand name were already coming. That traffic is cheap, it converts, and most of it would have arrived without the ad. It still belongs in your account, because you do not want a competitor buying that click. It just does not belong in the number you use to decide whether your advertising works.

The failure mode is quiet, which is why so many good brands sit in it for months. Total account ROAS looks healthy. You add budget. The new budget all lands on non-brand. Non-brand was already below break-even. The blended number barely moves because brand keeps carrying it. Six weeks later revenue is up and profit is down, and nothing in the dashboard told you.

Fixing it is a morning of work. Build a brand Search campaign on exact and phrase match around your name and your name plus a product word. Add your brand terms as an account-level negative keyword list so every other campaign, Performance Max included, has to leave them alone. Then pull a 90-day report with brand stripped out and read it again.

Two more structural rules. Never put brand and non-brand in a shared budget, because a shared budget feeds whatever spends fastest and brand always spends fastest. And name your campaigns so a stranger can read the account in 30 seconds: market, campaign type, brand or non-brand, margin tier. Future you is that stranger.

The map below is the shape most of our ecom accounts settle into. It is deliberately small. Every extra campaign splits your conversion data, and split conversion data is the thing that stops Smart Bidding from getting good.

If you only do one thing from this page, separate brand from non-brand and re-read your last 90 days. Most brands find the picture is less rosy and far more useful.

Go deeper: Brand campaigns for ecommerce, in full >

Lesson 3

How Do You Split Shopping and PMax by Margin Tier?

Tag every product with its margin tier in a custom label, then run exactly two Performance Max asset groups: high margin in one, everything else in the second with a tighter target. Across 12,304 Performance Max campaigns in our accounts, two asset groups beat one by 17 percent ROAS at the same spend, and five asset groups did worse than two. Two is the sweet spot because Smart Bidding needs enough conversions inside each group to learn.

Smart Bidding optimizes for the conversion value you send it, and the value you send it is revenue. It has no idea that your 40 euro necklace keeps most of its price and your 180 euro coat keeps a fifth of its price. Left alone it scales whatever converts most easily, and on most ecom catalogs that is the high-volume, low-margin end. Revenue climbs. Profit does not.

The fix is a feed field. custom_label_0 through custom_label_4 are free text columns you control. Put the margin tier in one of them. Three tiers is plenty: high, standard, thin. Then build your asset groups on that label rather than on whatever your store happens to call a collection.

A second label earns its place fast. Remember the concentration from lesson 1: 71.7% of products with ad spend sold nothing in their window. A label that separates proven sellers from products that have never converted is the quickest way to stop paying for the long tail without deleting it from the catalog.

Why not more groups? Because the data says more is worse. Five asset groups landed roughly 6 percent above the single-group baseline, well behind two. Every extra group divides the same conversions into smaller piles, and a group sitting under the volume it needs never learns. Two gives Smart Bidding a real difference to act on while keeping each side fed.

One warning on the split axis. Do not split by color, by supplier, or by product family unless that genuinely changes the bid you would want to make. The only splits worth making are the ones where you would honestly pay a different amount for the same click.

Two asset groups, split on a number that changes what you would bid. That is the whole finding from 12,304 campaigns, and it costs one afternoon of feed work.

Go deeper: What 12,304 Performance Max campaigns say about asset groups >

Lesson 4

Why Is the Product Feed Still the Biggest Lever?

Shopping and Performance Max have no keywords, so Google matches searches against your product data instead. Your feed is your keyword list. And the fields that decide the match are the ones almost nobody fills. In our Product Feed Quality Index, measured on 12,375 products across 251 real Merchant Center accounts on 2026-07-28, 81.3% of products carried neither a GTIN nor an MPN, and only 5% named a material or a pattern.

Read that index and one pattern jumps out. The fields your platform fills automatically are done everywhere: 97.5% of products have a brand and 96.2% have a condition. The fields a human has to sit down and fill are empty everywhere: material or pattern at 5%, product_detail entries at 10.8%, two or more extra images at 41%. Many of those accounts are measured at onboarding, before any optimization work starts, so the numbers describe how feeds arrive rather than how they leave. Either way, that gap is your opening.

Titles are the second opening. 68% of sampled titles run past 80 characters and the average store's average title is 91. Google shows around 70 before it cuts. So most brands are writing into space nobody reads, and burying the words buyers actually type behind their own product codes.

We also looked at 95,149 fashion titles from stores running Google Ads. The top 10% of products by revenue averaged 8.4 words and 50.6 characters. Products that sold nothing averaged 9.5 words and 56.3 characters. The clutter is specific: 18.3% of zero sellers carried a color word in the title against 12.2% of winners, and 45.1% carried a size token against 34.4%. Read that as correlation, because head products get hand-tuned titles while forgotten variants inherit auto-generated ones. The direction still holds. Lead with the product, and push color and size into the fields Google built for them.

In the same dataset, style words beat spec words. Indexed against their fair share of revenue, vintage came in at 1.12, floral at 1.11, cropped at 1.10 and linen at 1.06. Waterproof landed at 0.88 and UV400 at 0.89. In fashion people buy a look, not a spec sheet. Write the title accordingly.

Two hard rules while you work. Never invent a GTIN. If the manufacturer did not give you one, leave it empty and set the identifier flag honestly, because a made-up code is a fast suspension for ten minutes saved. And never let a feed tool overwrite product_type. That is your own category path, and your campaign splits are built on it.

Spend a week on titles and attributes before you touch a bid. The feed is the only lever that keeps paying after you stop working on it.

Go deeper: The Product Feed Quality Index, with the full data >

Lesson 5

What Does Search Do When Performance Max Already Runs?

Search does three things Performance Max cannot. It owns your brand name. It holds exact position on the phrases you must win. And it shows you the search terms Performance Max keeps to itself. Run brand Search on its own, run a tight non-brand campaign on exact match for your best category and hero product phrases, and mine the search terms report every week to feed negatives back into the whole account.

Google decides between your own campaigns before it decides between you and a competitor. A Search campaign with an exact keyword match generally takes priority over Performance Max for that query. So Search is how you keep control of the terms you care most about, instead of hoping PMax bids the way you would have.

Performance Max reporting is also thin by design. You can get asset group data and, with some work, product-level data. You cannot get a full search terms list. A Search campaign running on the same themes is your window into what people actually type in that space, and every negative you build from it applies account wide.

Keep it small. On an established brand you are not rebuilding the catalog as keywords. You want your category terms, your hero products, your brand plus modifiers, and that is mostly it. Exact match on the phrases you must win, a small phrase-match budget for discovery, and broad match only if you have a strong negative list and value-based bidding underneath it.

The weekly loop is fifteen minutes and it is the cheapest improvement in the account. Open the search terms report, sort by cost, and look at anything that spent money without producing a sale. If a term does that two weeks running, it becomes a negative. Apply the list at account level so Performance Max has to obey it too.

Competitor bidding is a decision, not a drift. It is allowed, it is expensive, and it only pays when your offer is genuinely stronger on the thing they are searching for. Give it its own budget and read it on new customers, never on blended ROAS.

Performance Max is the engine. Search is the steering and the windscreen. Running one without the other is why so many accounts feel like a black box.

Go deeper: Performance Max vs Search campaigns, side by side >

Lesson 6

Which Assets Actually Matter in Performance Max?

Start feed-only, then add assets to a copy of the group that already works. Feed-only Performance Max keeps your budget on Shopping-style placements where buying intent lives. When you do add assets, the ones that move an ecom account are clean product photography, two or three real lifestyle shots, one short vertical video, and headlines written from your own search terms. Upload every size Google asks for, or it will crop your product out of frame.

Feed-only means an asset group with no headlines, images or video, just products. Google keeps it close to Shopping. That is the right start on a store that already has product demand. It stops your budget spraying across YouTube and Display before you know what a good cost per sale even looks like here.

Add assets when the feed-only group is consistently profitable and you want more reach than Shopping placements can give. Add them to a duplicate of the group rather than the working one, so when the numbers move you know which change moved them.

If you do not upload a video, Google generates one from your images, and it will be worse than anything you would ship yourself. A 10 to 15 second vertical clip of the product being used beats an auto-generated slideshow every time, and it is a phone and a tripod, not a production.

Write headlines out of your search terms report and the language in your own reviews. On an established brand you already know the phrases people use. Do not put brand poetry in a slot Google will render next to a price and a shipping estimate.

The honest bit: creative matters far less on Google than on social feeds. On Meta the creative is the targeting. On Google the feed is the targeting, and the assets mostly decide whether the click happens once you have already won the match. If your assets are perfect and your titles are 91 characters of product code, you are polishing the wrong thing.

Feed first, assets second. On Google the picture that sells is the one in the Shopping card, and that comes from your feed, not from your asset group.

Go deeper: Performance Max best practices for ecommerce >

Lesson 7

How Do You Know Your Google Ads Numbers Are Real?

Send one purchase conversion action carrying the real order value, mark only that one as primary, switch on enhanced conversions, add server-side tracking, and reconcile against Shopify every week. Pixel-only setups lose a large share of sales to browser privacy rules, and server-side tracking typically recovers 30 to 40 percent of what the pixel misses. Google bids on what it can see, so a hole in the data is a hole in the bidding.

Shopify is the source of truth for what your store actually sold. Google Ads tells you what Google can attribute. Those two numbers will never match exactly and that is fine, but you need to know the size of your normal gap before you read a single ROAS figure. When the gap moves, something broke, and it is nearly always tracking rather than performance.

Two breaks account for most of the damage. The first is double counting: a Google Ads purchase tag AND an imported GA4 purchase, both marked primary. Every sale counts twice, ROAS looks superb, you scale, and the bank account goes the other way. The second is a fixed conversion value. If every order reports as 1, Google cannot tell a 30 euro sale from a 300 euro sale, so value-based bidding has nothing to work with.

The learning phase is the other thing established brands break constantly, because they have the budget to keep tinkering. Every material change restarts learning: a budget move over roughly 20 percent, a new target ROAS, a structural edit, a new asset group. Expect one to two weeks where the numbers are not a verdict. Write down what worries you, wait it out, then decide with data.

Attribution deserves one decision and then silence. Purchases in fashion, jewelry and home decor are not same-day, so a 30-day click window with data-driven attribution reads very differently from last click on a 7-day window. Neither is wrong. Switching between them mid-quarter is wrong, because it makes your own history unreadable. Pick one, write down the date you picked it, and leave it alone.

1. Step 1

   One purchase action, carrying real values

   Count it as one per click and pass the real order total dynamically from the order. This single action is what your whole account gets judged on, so it has to be the only thing Google is chasing.

2. Step 2

   Everything else goes secondary

   Add to cart, begin checkout, newsletter signups. Useful for reporting and audiences, poisonous as a bidding target. Two primary purchase actions means every ROAS number in the account is wrong.

3. Step 3

   Turn on enhanced conversions

   It sends a hashed, privacy-safe version of the customer email with the conversion so Google can match sales it would otherwise drop. It is free, it takes minutes, and it recovers real data.

4. Step 4

   Add server-side tracking

   Order data goes from your Shopify back end straight to Google, so it survives ad blockers, Safari privacy rules and the next browser update. This is what closes most of the 30 to 40 percent gap.

5. Step 5

   Pick one attribution model and one window

   Data-driven attribution on a 30-day click window is the default we run on ecom accounts. What matters more than the choice is that you stop changing it and compare like with like.

6. Step 6

   Exclude what you do not want more of

   Wholesale orders, staff orders, subscription renewals that were always going to happen. If they land in the same purchase event, Google will happily go and find you more of them.

7. Step 7

   Reconcile against Shopify every week

   Compare Google Ads conversions to real Shopify orders from Google traffic. Learn your normal gap. Close any new gap before you touch a single bid or budget.

Prove the numbers are complete before you touch a bid. Most of the accounts where the algorithm looks broken turn out to be accounts where Google never saw a third of the sales.

Go deeper: Google Ads conversion tracking on Shopify, set up properly >

Lesson 8

How Do You Add Budget Without Tanking ROAS?

Raise budget in 20 to 30 percent steps, hold each step for a full week, and check the marginal step rather than the account average before the next one. Know your break-even ROAS first: it is 1 divided by your gross margin. A store keeping 60 percent gross margin breaks even around 1.67. A store keeping 30 percent breaks even at 3.33. The same 3x is a win for one and a slow loss for the other.

Gross margin is what is left of a sale after the cost of the goods, as a share of the sale price. Sell for 100, the product cost you 40, gross margin is 60 percent, break-even ROAS is 1 divided by 0.60, about 1.67. That is the floor where the ads pay for the goods and nothing else. Shipping, payment fees, returns, apps and people all sit below that line, so your real target has to be meaningfully above break-even, not next to it.

Big jumps hurt for two reasons at once. They reset the learning period, and they force Smart Bidding to buy cheaper, worse traffic to spend the money. You then read the bad week as proof the campaign is maxed out, when what you actually measured was the jump.

On an established brand the next euro almost never belongs in a new campaign. Check in this order. Is a proven product losing impression share purely on budget? Is a high margin product capped by a target that is too tight? Is stock the real limit? Only after all three does a genuinely new market or campaign type make sense.

Every account has a point where more budget only buys worse traffic. You find that point by walking up the staircase and watching the marginal step, never by looking at the blended number. If the last 20 percent of budget added sales below break-even, that step was a donation, even though the account average still looks perfectly healthy.

Check stock cover before every step. Scale into a product with three weeks of cover left and you spend a fortnight teaching Smart Bidding to love it. Then it sells out, and you hand the learning straight back.

Scale on the marginal step, not the account average. The question is never whether your ROAS is good. It is whether the last 20 percent of budget earned more than it cost.

Go deeper: How to scale Google Ads for ecommerce >

Lesson 9

How Do You Plan Q4 Without Breaking the Account?

Do the structural work in September, then stop restructuring. Feed fixes, margin labels, asset group changes and new campaigns all need learning time, and learning time in November is the most expensive thing you can buy. From October the only levers you touch are budget, targets and which products you push. Everything else is frozen until January.

A structural change in peak week costs you the learning period at the exact moment traffic is worth the most. Auction pressure is already higher, so your cost per click is higher, so the week you spend relearning costs several times what the same week costs in June. That is the whole argument for the freeze.

Loosen the target ROAS before peak, not during it. Conversion rates rise in Q4, which means a target that was correct in September quietly turns into a brake in November: the campaign clears it easily and then stops spending. Give it room a couple of weeks early so the change has settled before the traffic arrives.

Season is not only a Q4 thing. Category demand rotates faster than most brands adjust for. We track 134 store-level trends, 112 fashion accounts and 22 home decor accounts. Into July, outerwear shed 2.5 points of revenue share while dresses gained 1.5 and bottoms gained 1.1. Bottoms were climbing in 56.8 percent of the stores that sell them. If your labels and budgets still reflect last season, you are paying to push what the market already moved on from.

Put your real shipping cutoff dates in the feed and in the ads. Selling a delivery date you cannot hit turns a great December into a January of refunds, chargebacks and a Merchant Center problem you did not need.

January is the other half of the plan and the half everyone skips. Conversion rates fall back, your Q4 winners are out of stock or out of season, and the labels you set in September are now wrong. Rebuild the tiers in the first week of January instead of watching the account bleed quietly through the month.

Q4 is won in September. By November your only honest levers are money and stock.

Go deeper: The Black Friday and Q4 playbook for ecommerce >

Lesson 10

When Should an Ecom Brand Hire Google Ads Help?

Hire when the account already makes money and your own time is the thing holding it back. If nobody in the business can protect a focused half day a week for the feed, the search terms and the labels, that work does not happen, and that work is where the money is. If the account loses money because the product or the margin is wrong, an agency will not fix that, and any agency promising it should worry you.

Straight disclosure so you can weigh this properly: we are a Google Ads agency for ecom brands. We have written the honest version anyway, including the cases where hiring anyone is the wrong call.

Keep it in house while you are learning the mechanics, while the spend is small enough that a mistake is cheap, or when someone internal genuinely enjoys the detail work. The free Google Ads eCom Lab exists for exactly that stretch: the course is inside, calls run monthly, and operators answer real account questions in chat. Learning it yourself also means you will understand every report anyone ever shows you afterwards.

If you do hire, the things worth checking are boring and checkable. Who actually touches the account day to day, not who is on the sales call. Whether the fee is a flat retainer or scales with spend. Whether they will work the product feed or only the campaigns, because the feed is where most of the money is. And what happens to your Merchant Center, your labels and your tracking if you leave.

Expect the first 60 days of any handover to be repair rather than growth: feed, tracking, structure, labels. Anyone promising a step change in week one is either restructuring recklessly or quietly reading your brand traffic back to you as their own result.

And be honest about the cases nobody can help with. A thin margin, a product nobody searches for, or a store that converts at half the category norm are business problems wearing a marketing costume. Paying a percentage of spend to have them managed is an expensive way to find that out.

The trigger is your calendar, not your revenue. When the highest-value hour in your week is feed work you keep not doing, it is time.

Go deeper: What a Google Ads agency costs for an ecom brand >