How to Scale Your Store Properly

How Google Ads makes a store money

Google will happily spend your whole budget on sales that lose you money. It's doing exactly what it was told. The problem is who told it, and what.

Strip away the dashboards and Google Ads does one thing: it auctions the attention of people who are looking for something. Every time someone searches, scrolls YouTube or opens Gmail, Google decides in milliseconds whose ad to show and what that click will cost. You don't pay for being seen. You pay when someone clicks.

For an online store, almost all of that happens in three places:

The part almost everyone misses

Modern Google Ads is automated. You don't set each bid yourself anymore. You tell Google what a sale is worth, and it bids to get you more of whatever you said was valuable. That means the most important setting in your account isn't a bid, a keyword or a budget. It's the number you send back after each purchase.

Out of the box, that number is the order total. So Google learns to chase revenue: big baskets, popular products, easy sales. It has no idea that your most popular product makes you $2 after shipping, or that one in five of a certain item comes back.

Google doesn't know what you make. It only knows what you tell it. Tell it revenue, and it buys you revenue. Even at a loss.

The loop every account runs on

1. Data in: your product feed, your conversion tracking, the value of each sale.
2. Bids out: Google decides who to show, where, and how much to pay.
3. Results back: sales get reported, and Google adjusts.

Almost every "my ads stopped working" story is a problem in step one. Bad data in means confident, automated, expensive decisions out. That's why this playbook starts with your numbers and your tracking, not with campaigns.

So what number should Google be chasing? First you need to know your own. That's next, and it takes five minutes.

Your numbers before your ads

Two stores spend $10,000 a month at a 4.0 ROAS. One keeps $6,900 after the ads are paid. The other loses $1,000. Same dashboard. Same green numbers. Here's the one that tells them apart.

Before you can tell whether ads work, you need three numbers. Most store owners know their revenue to the penny and these three only roughly. That gap is where the money leaks.

1. What one order actually leaves you

Not the order value. What's left after the product, the shipping you pay, payment fees and the orders that come back. Call it profit per order before ads.

Returns are the one people forget. If 10% of orders come back, you lose that revenue, you often pay shipping both ways, and some stock can't be resold.

2. Break-even ROAS

ROAS (return on ad spend) is revenue divided by ad spend. A ROAS of 4 means $4 of sales for every $1 of ads. It sounds healthy. Whether it is depends entirely on your margin.

If an order is worth $60 and leaves you $20 before ads, your break-even ROAS is 3. Below 3 you're paying to be busy. Above 3 you're making money. Two stores can both show ROAS 4: one is printing money, the other is losing it.

3. POAS: the number this whole playbook is built on

POAS (profit on ad spend) asks the question ROAS doesn't: for every $1 the ads cost, how much profit came back?

A POAS of 1.3 means every $1 of ads brought back $1.30 of profit: you kept 30 cents after paying for the ad. A POAS of 0.8 means you lost 20 cents on every dollar, however good the ROAS looks.

Do this per product, not just per store

Your store average hides the truth. A 60% margin candle and a 15% margin furniture piece can't share a target. Pull your top 20 products by ad spend and work out the profit per order for each. You'll usually find a few heroes paying for a long tail that isn't. Chapter 3.2 turns that into a system.

Got your POAS? Good. Now the bad news: it's only as true as your tracking. Next chapter: the check that catches the lies.

Tracking that tells the truth

One store's Google Ads showed 8x more sales than the store actually took. The owner was scaling on it. Every automated bid was spending real money chasing sales that never happened.

Automated bidding is only as good as what it's told. If Google counts sales twice, misses a third of them, or never hears about refunds, it will bid with total confidence on the wrong things. Fixing tracking is the cheapest growth you will ever buy.

The weekly gut check

Once a week, compare two numbers: orders in Shopify, and purchase conversions in Google Ads, for the same seven days. They will never match exactly. Google only claims orders that came after an ad, and attribution windows differ. But they should move together.

Get these five right

1. One primary purchase action. In Goals → Conversions, only one purchase conversion should be set as primary. Everything else (add to cart, sign-ups, a GA4 import) is secondary, so it's reported but not bid on.
2. The value is real. At minimum, the order value excluding tax and shipping. Later (Chapter 3.1) you'll send profit instead.
3. Enhanced conversions are on. They let Google match a sale to an ad click using hashed customer data, which recovers sales that cookies lose.
4. Consent mode is set up if you sell in the UK or EU. Without it, Google can't model the shoppers who decline cookies, and those sales go missing from what it learns.
5. Refunds reach Google. If a returned order still counts as a sale, your ROAS looks better than it is. Conversion adjustments let you retract or restate refunded orders.

For most Shopify stores, the Google & YouTube app connects the purchase tag and the product feed in one place. It's a good default. What matters is not which tool fires the tag, but that you know exactly which action Google is bidding on, and what value it sends.

If you don't trust the number, neither should Google. Fix tracking before you touch a single bid.

Tracking fixed means Google can finally learn. But learn what, about which products? That's decided by a file most stores haven't opened since day one.

Merchant Center and your product feed

On Shopping there are no keywords. Your product title is your targeting. Which means most stores are fighting for shoppers with titles like "The Everyday".

Shopping and Performance Max don't use keywords. Google reads your product feed, the list of every product with its title, description, image, price and attributes, and decides which searches to show each product for. Your feed is your targeting. A lazy feed is a lazy account, no matter how good the campaigns are.

Titles do the work keywords used to do

Google matches searches mostly against the title. Put the words a shopper types first, and your brand last unless people search for it.

A reliable shape: product type + key attribute + material or size + who it's for + brand. Start with the 20 products that get the most spend, not all of them.

What to check every week

• Needs attention: disapproved or limited products. Every flagged best seller is sales you're not even competing for.
• Price and availability match the site. A mismatch gets products pulled, often in your busiest weeks.
• GTINs (barcodes) on branded products, so Google can match you to the right searches.
• Images: clean product shots first. Lifestyle images can go in as additional images.

The two fields that set up everything profitable later

Cost of goods. Merchant Center accepts a cost_of_goods_sold value per product. With it, Google can report gross profit per product, not just revenue. Add it now, even if you don't use it yet.

Custom labels. Five free fields (custom_label_0 to 4) that you fill with anything you want to group by. This is how you'll later split products by margin, so a 60% margin product and a 15% one never share a target:

Your feed is your targeting. Most accounts we open have a feed nobody has touched since the day it was connected.

Your foundation is done. Level 2 builds the campaigns on top of it, structured so a 60% margin product and a 15% one never fight over the same budget again.

Search: brand, non-brand and negatives

The best-looking campaign in most accounts is the one doing the least work. It's the people who already typed your name.

Search ads show when someone types a query. Store searches split into two kinds, and they behave like two different businesses:

• Brand: people searching your store or product name. They already know you. Clicks are cheap and the ROAS looks incredible, because many of them would have bought anyway.
• Non-brand: people searching the problem or the product ("orthopaedic insoles", "plantar fasciitis shoes"). These are new customers. This is where growth comes from, and where money gets wasted.

Split them, always

Run brand and non-brand as separate campaigns, with separate budgets and separate targets. Then add your brand name as a negative keyword in every non-brand campaign, so the cheap brand clicks can't hide inside the expensive ones.

Non-brand: start narrow, earn the right to go broad

1. Start with buying searches: the product plus a problem, a size, a type or "buy". Leave "what is" and "how to" for content, not ads.
2. Group by intent, not by your menu. "Insoles for heel pain" and "insoles for flat feet" are different shoppers and deserve different ads and landing pages.
3. Match types: exact and phrase for your proven best terms. Broad match only with automated bidding running on reliable conversion data (chapters 1.3 and 2.3), or it will find every loosely related search there is.

Negatives: the cheapest money in the account

Every week, open the search terms report and read what you actually paid for. You will find people looking for jobs, free stuff, repairs, competitors' stores and things you don't sell.

Put the repeat offenders in a shared negative keyword list and attach it to every campaign: free, jobs, cheap, DIY, second hand, repair, PDF, used. Add your own as you find them.

Brand campaigns make reports look good. Non-brand campaigns make the business bigger. Never let one hide the other.

Search finds the people who type. But for most stores, the money is in Shopping, and that's where structure decides everything.

Shopping and Performance Max, structured by margin

Most Shopping accounts are built like the store menu: shoes, socks, accessories. Google doesn't care about your menu. Your bank does care about your margins.

Here's the problem with structuring by category. A product that keeps 55% of the price and one that keeps 20% end up in the same campaign, with the same ROAS target. Google hits the target by selling more of whatever sells easiest. Often that's the low-margin stuff.

Group by what each product keeps

Remember the margin label from chapter 1.4? This is where it pays off. Every margin group gets its own break-even, and its own target:

Standard Shopping or Performance Max?

Our rule: prove it in Shopping, scale it in Performance Max. Turning on Performance Max first, before you know which products make money, means handing a black box your budget and hoping. Once a margin group runs above its break-even in Shopping, give it its own Performance Max campaign or asset group with a target to match.

Category structure tells Google what you sell. Margin structure tells Google what you earn. Only one of them makes you money.

Now each group has a target. But set a ROAS target too early and Google simply stops spending. The order you unlock bidding in matters.

Bidding, step by step

Automated bidding isn't a switch. It's a staircase. Skip steps and Google either stops spending or spends on anything that moves.

Automated bidding learns from conversions. With little data, it guesses. With good data, it's better than any human. Your job is to give it the right goal at the right moment.

1. New campaign, no sales yet: maximise clicks with a cap, or manual bids. You're buying data.
2. Sales are coming in: maximise conversions. Google learns who buys.
3. Sales with real values: maximise conversion value. Google learns who buys more.
4. Roughly 30+ conversions a month in that campaign: target ROAS, set at or a little above break-even. Now it has to buy value at a price.
5. Profit is the conversion value (chapter 3.1): the same target ROAS now means target profit. This is the end of the staircase.

The three rules that stop you breaking it

• Move targets 10–20% at a time. Big jumps reset the learning and spending swings wildly.
• Wait one to two weeks between changes. Judging a change after three days is reading noise.
• Never judge a campaign while it's learning. The status column says "Learning". Leave it alone.

Most "Google Ads stopped working" moments are someone changing three things on a Tuesday, then two more on Thursday.

Targets set. Now the two things everyone gets wrong last: how much to spend, and what the ad actually says.

Budgets and ad copy that sells

A budget is not a goal. It's a ceiling. The right budget is however much your profitable target can actually spend.

Budgets: let profit set the number

• Budget follows the target, not the other way round. If a campaign holds its target ROAS and still spends the whole budget by mid-afternoon, it's leaving profitable sales on the table. Raise it in steps.
• If it can't spend the budget at its target, that's Google telling you there aren't more sales at that price. Forcing more money in means lowering the target, and lower targets mean thinner profit.
• Budget-limited campaigns behave differently. Since a Google change in August 2026, target-ROAS Shopping campaigns that are limited by budget aim at the target instead of beating it. Watch the "limited by budget" status.
• Put the most money where the margin is. High-margin groups first, low-margin groups last.

Ad copy: say what the shopper is thinking

Responsive search ads take up to 15 headlines and 4 descriptions, and Google mixes them. Feed it variety. Every headline should earn its place with one of these five jobs:

Use only claims that are true on your site today: your real review count, your real returns policy. Then add assets: sitelinks to your best categories, a promotion if there is one, and price assets for your top products.

The ad doesn't need to be clever. It needs to sound like the answer to the exact thing they just typed.

Level 2 gives you an account built around margin. Level 3 goes further: Google stops bidding on revenue at all, and starts bidding on profit.

Feeding Google profit instead of revenue

This is the chapter the whole playbook was building to. Change one number, and every automated bid in your account starts working for your bank balance instead of your revenue.

Remember chapter 1.1: Google bids to get you more of whatever value you send after a purchase. Today that's almost certainly the order total. Here's what changes when you send profit instead:

Three ways to do it, simplest first

1. Margin-weighted value. Send the order value multiplied by that product group's margin share. Rough, but a big step up from raw revenue.
2. Real profit per order. Calculate revenue minus product cost (and ideally shipping and fees) for each order, and send that as the conversion value. Usually done server-side or through a profit-tracking tool connected to your store.
3. Cost of goods in Merchant Center, plus cart data. With cost_of_goods_sold in your feed (chapter 1.4) and cart data in your purchase tag, Google can report gross profit per product. Use it to see, then use option 2 to bid.

Switch carefully

• The value scale changes, so the target changes. A target ROAS of 3.0 on revenue might become a target of about 1.2 on profit, meaning $1.20 of profit for every $1 of ads.
• Switch one campaign first, ideally a steady one. Expect one to two weeks of learning.
• Keep a revenue conversion running as secondary, so you can still compare.

Every store says it wants profit. Almost every account tells Google it wants revenue. This chapter closes that gap.

Google now sees profit. Time to see it yourself, product by product, because some of your best sellers are about to look very different.

Product triage: heroes, villains and zombies

Somewhere in your catalogue is a best seller that loses you money every time it sells. It's probably getting more budget than anything else.

Every product you advertise falls into one of four groups. The job is to find out which, and treat each one differently:

Your bestseller list is a revenue list. Your hero list is a profit list. They are rarely the same products.

Villains are bad. But nothing creates villains faster than a 30% discount with the old targets still running.

Returns and discounts: the maths before Black Friday

A 30% discount on a 60% margin product doesn't cut your profit by 30%. It cuts it in half. And your ad targets usually don't change at all.

A discount comes straight off your margin, not your revenue. That's why sales feel busy and pay badly. Before any sale, run the numbers:

Before every sale

1. Raise the ROAS targets on discounted products to the new break-even, for the length of the sale.
2. Prefer offers that cost less than they look: a free gift over a threshold, free shipping, or a bundle. These usually protect margin better than a flat percentage.
3. Exclude your lowest-margin products from the sale, or from ads during it.

Returns: the silent discount

A 20% return rate is a 20% discount you never chose. Count it. Feed refunds back to Google as conversion adjustments (chapter 1.3) and use your real return rate per product in the triage tool. In fashion and footwear especially, a product with a high return rate can be a villain hiding behind a great ROAS.

Discounts don't hurt because people buy less. They hurt because nobody changed the targets.

Margins protected. Now, finally, the part everyone wants to skip to: spending more.

Scaling in steps, only where profit holds

Here's why your ROAS drops every time you push the budget. Your next dollar is never as good as your average dollar.

Google spends your budget on the best opportunities first. Every extra dollar buys a slightly worse one. So your average POAS can look fine while the last few thousand dollars are already losing money.

The step-scale method

1. Only scale campaigns that sit above target on profit, not ROAS.
2. Raise budget by 15–20%. Leave the target alone.
3. Wait one to two weeks.
4. Check what the extra spend brought in: extra profit divided by extra spend. Above 1.0, take the next step. Below 1.0, step back to the last level and hold.

Scaling isn't spending more. It's spending more only while the next dollar still pays you back.

Every chapter so far needs one habit to stick: looking at the right numbers every week. Here's the page.

The weekly profit page

If your weekly report doesn't fit on one page, nobody reads it. If it doesn't show profit, it isn't a report.

Every Monday, one page. Same layout every week, so changes jump out:

Example figures, to show the layout.

Under the numbers, four short lists

• By margin group: spend and POAS for high, mid and low.
• Top 5 heroes and top 5 villains from the triage (chapter 3.2).
• Changes made this week, with the reason for each.
• One decision for next week. Just one.

Revenue reports tell you what happened. A profit page tells you what to do on Monday.

And when the page turns red week after week? Here's how to find out why, in the right order.

Diagnosing a dying account

When results drop, the instinct is to change the campaigns. The cause is almost never the campaigns. Check in this order, and stop at the first thing that's broken.

1. Tracking (chapter 1.3). Did the Shopify vs Google gap jump? Fix this first or nothing else can be judged.
2. Feed (chapter 1.4). Disapprovals, price mismatches, a best seller out of stock.
3. Store. A slower page, a theme update, a checkout change, a price change.
4. Market. A competitor running a sale, the season turning, a new player bidding on your terms.
5. Only then, the account. Recent changes, targets moved too far, budgets limited.

If someone else runs your ads: ask them these ten

Most dying accounts aren't dying. They're being fed bad numbers, and nobody checked the numbers first.