The Ecommerce Google Ads Account Structure That Actually Scales

Most stores hit a wall when they try to scale. Budgets go up, ROAS goes down, and nobody can say why. The reflex is to blame the bidding, or the creative, or the algorithm having a bad week.

It’s almost never that. After auditing a lot of ecommerce accounts, the pattern is boring and consistent: the structure that worked at $3k a month was never built to hold $10k. Scaling didn’t break the account. It exposed it.

So this is the structure I build toward, laid out as eight building blocks. Every block is a job. Branded search has one job. Shopping has another. Performance Max has another. When each campaign knows its job, you can read the account, and an account you can read is an account you can scale.

The blocks aren’t a menu you pick from. They switch on in an order, across three phases.

The three phases everything sits on

Before the blocks, the timeline they live on:

• Phase 1, Foundation. Clean data and clean separation. Tracking that isn’t double-counting, branded split out from everything else, five or fewer campaigns total. You’re not trying to win yet. You’re trying to give Google honest signals and give yourself a dashboard you can trust.
• Phase 2, Efficiency. Get each campaign to enough volume to actually learn (think 50+ conversions a month), then push budget toward what’s already working and starve what isn’t.
• Phase 3, Scaling. Grow budget and reach on proven campaigns, on purpose, with rules, so growth doesn’t turn into a slow profit leak.

You can’t skip ahead. A Phase 3 move on a Phase 1 account is how people light money on fire and decide “Google doesn’t work.” Here’s how the eight blocks land on that timeline.

1. Start with the goal, not the ROAS

This is the block everyone skips, and it decides whether the other seven mean anything.

If your only target is Google Ads ROAS, you’ve picked a number that can’t tell you if the business is growing. Platform ROAS ignores your organic and direct sales, ignores assisted conversions, and has no real connection to profit. You can grow that number to the moon by leaning on brand and remarketing while new-customer revenue flatlines.

Pick metrics that describe the actual business instead:

• MER (blended ROAS): total revenue divided by total ad spend, across every channel. This is the scoreboard. I wrote a full breakdown in why blended ROAS is the metric to obsess over.
• New customer acquisition. Google over-indexes on the bottom of the funnel, so it loves selling to people who were going to buy anyway. Growth comes from new customers, so measure them.
• Contribution margin. A 4x ROAS can still lose money. Margin, not ROAS, tells you if a sale was worth making, which is the whole point of running margin analysis.
• Product mix. Adding a flood of low-AOV, low-repeat products can reshape the whole account, usually for the worse.

You can’t grow profit, MER, and new customers all at once without trade-offs. Choose what you’re optimizing for first. Then give every campaign a single, stated job, so when you look at the account you know what each line is supposed to be doing.

2. Branded search: protect it, don’t lean on it (Phase 1)

People searching your brand are warm. You want those sales, you want to keep competitors off your name, and broad match means rivals can show on your brand terms even when they’re not bidding on them directly. So yes, run branded search.

Two rules keep it honest.

First, it lives in its own campaign and gets reported separately. The moment brand is mixed into other campaigns, you’ve blinded yourself to how much real, new demand the account actually creates.

Second, don’t confuse a gorgeous brand ROAS with growth. Brand mostly captures demand that already exists. If you let a smart bidding strategy keep pushing more spend onto your own name, the ROAS stays beautiful while the account stops growing. Brand can eat 50 to 80% of a budget if you let it, and that spend is harvesting, not hunting. Cap it, watch your impression share and auction insights, and keep brand as a small, protected slice.

3. Shopping is the foundation (Phase 1 into Phase 2)

For most ecommerce accounts, Shopping drives the most new-customer sales of anything except brand. Get it right and the account has a floor. Get it wrong and everything above it wobbles.

Start simple. One clean Standard Shopping campaign with all your products in it. A single well-fed Shopping campaign, run long enough, is hard to beat, and it gives smart bidding the volume it needs.

What actually moves Shopping isn’t a clever bid trick. It’s the feed. The most important file in your account isn’t even in Google Ads, it’s in Merchant Center. Get all products approved, no disapprovals, keyword-rich but natural titles, correct product types, good images and prices, and import your cost of goods at the SKU level so you can manage by margin. Google is good at understanding products now, so don’t agonize over a feed for weeks, but don’t ship a weak one either.

Then, and only then, structure.

The five Shopping structures (and the one rule for splitting)

Most accounts I audit have too many Shopping campaigns. Seven, ten, twenty. And not one of them has enough conversions for smart bidding to learn. Splitting feels like control. It’s usually just starvation, because each campaign needs roughly 30 conversions a month before the bidding has anything to work with. I’ve taken an account from over 20 campaigns down to 5 and watched performance improve, simply because the survivors finally had volume.

So the golden rule: make one campaign profitable before you split anything. If you can’t make a single campaign work, no amount of clever segmentation saves you.

Once it’s working, here are the five structures worth using:

1. One-campaign baseline. All products, one campaign. Start here, always.
2. Bestsellers + Main. Pull proven winners into their own campaign so you can fund them harder.
3. Main vs Secondary product lines. Split when two lines genuinely behave differently.
4. Promotional vs Full price. Isolate sale periods so discounts don’t distort your everyday economics.
5. Seasonality. Separate seasonal ranges so they can be scaled up and wound down cleanly.

Pick one. Don’t stack them. And only split when products have meaningfully different conversion rates, or you need ROAS targets at least 20% apart. Validate the split by conversion rate, not by ROAS. Simple and profitable beats complex and mediocre every time.

I walked through the current version of all five in a recent video:

4. Performance Max: guardrails, or it eats your brand (Phase 1 into 2)

Performance Max can carry an account, and it can hollow one out. The difference is guardrails.

Left alone, PMax chases the easy conversions. It piles onto branded searches and cheap remarketing, posts a dreamy ROAS, and tempts you to raise the budget. Do that and new-customer volume slowly drains, sales follow, and the ROAS that looked so good was the warning sign.

Set the guardrails:

• Exclude your brand. Use brand exclusions, and the newer setting that lets you keep brand out of Search while still allowing it on Shopping, so PMax does real Shopping work instead of reselling your name.
• Watch where the money goes. PMax stopped being a black box. The channel and search-terms reports show you how much is going to Search, Shopping, Video, and Display, and what each returns. If the spend is mostly brand, Gmail, and Display, you’re not growing. More on the auto-generated side of this in the ads PMax writes that you never wrote.
• Track new vs returning. This is the metric that tells you whether PMax is acquiring or recycling.
• Feed-only vs full assets is a test, not a belief. Run both asset groups on the same products, give it a week, and if Video and Display are eating more than half the budget, pause the full-asset group. Give it a month, and if the non-Shopping ROAS misses your target, lean on feed-only. Let the data choose.

One change worth knowing: PMax no longer automatically wins ties against Standard Shopping. They compete on Ad Rank now, so a well-run Standard Shopping campaign can hold its ground.

If an account is thin on conversions, feed it micro-conversions (add-to-cart, begin checkout) so the bidding has signals to learn from while real purchases build up.

5. Non-branded search: seed it from data (Phase 2)

Text search for non-brand terms can drive real volume, but it’s where people overbuild the earliest.

Don’t open with a big keyword build. Let Shopping and PMax run first, then harvest the search terms that actually converted and turn those into keywords. Start them on exact match, pointed at the right product or collection page. Add AI Max or a controlled broad-match layer later, once you have proof. These are proven terms going to proven pages, so they convert, but the volume can tap out fast, so don’t expect search alone to carry growth.

The exception is restricted categories. If you’re in supplements or anything that gets bounced around in Merchant Center, search may have to do more of the heavy lifting, so weight it accordingly.

6. Demand Gen and YouTube: careful, mostly Phase 3

This is your top of the funnel inside Google, creating demand rather than capturing it, much closer to a social ad than a Shopping ad. It can work, and it’s almost entirely creative-driven: the image and video carry the result, and the attribution window is long unless the offer is sharp.

It’s also where I see the most waste. People pour budget into Demand Gen while the rest of the account is profitable and this part isn’t. Be honest about the trade. A dollar that could have gone to Shopping or PMax usually does more there, and if you want pure prospecting, social often beats it. At a minimum, use it to remarket, reuse the creative you already feed PMax, and don’t let it balloon. Earn it in Phase 3, don’t lead with it.

7. Budget allocation: don’t split it too thin (every phase)

Budget is how you tell Google what matters. More budget on a campaign is an instruction to focus there. So allocate on purpose.

Set a rough percentage per campaign type and resist slicing it into too many pieces, because every campaign still needs enough volume to learn. Use shared budgets for brand so it can’t run away. Push budget toward campaigns that are constrained and driving new customers, and pull it from anything dragging the account down. Often the move isn’t more budget at all, it’s fixing what’s capping the campaign you already fund. I covered that in getting unstuck when you’re limited by budget.

8. Scaling rules: the part everyone skips (Phase 3)

Most people scale with no rules. ROAS ticks up, they raise the budget, it gets away from them, profit evaporates, and they conclude Google is broken.

Scaling is a balancing act, and it needs a written rule set specific to your business:

• An MER floor and a minimum profitability threshold you won’t cross.
• New-customer targets, so growth means new buyers and not just more brand harvesting.
• Budget increases in deliberate steps (10 to 20% at a time), so you can see the effect before the next one.
• tROAS loosened gradually. Set it too tight while scaling and you choke the campaign.
• A standing rule that you never scale on in-platform ROAS alone. Watch top-line sales, margin, and what actually lands in the bank.

This is where the three phases pay off. You only earn a Phase 3 move once Phase 1 and Phase 2 hit their marks: tracking clean, brand separated, five-or-fewer campaigns to start, then 50+ conversions a campaign at a healthy MER. Build the structure first. Then scale it.

Build the floor, then raise the ceiling

None of these blocks is exotic. The edge is in the order. Goal before campaigns. Foundation before efficiency. Efficiency before scaling. An account built this way is one you can actually read, and the readable account is the one that survives going from $3k to $30k a month.

If you want the whole thing on one page, I keep a 3-phase growth framework you can work through, and a free shopping and PMax scaling guide that goes deeper on the when-and-how-much of raising budgets.

So before your next budget increase: does every campaign in your account have one clear job, or are you about to scale something you can’t read?