How I run Google Ads accounts for eCommerce: the system
Your Google Ads account is either a system or a slot machine.
This is the full method I use on every account. Every decision, every sequence, every tool, in the order I actually do them. Read it before we talk. If it matches how you think about your business, we should work together. If it does not, you will know before either of us loses time.
Short version: profit before ROAS, tracking before bidding, feed before campaigns, structure that matches your spend, and three sentences every Friday instead of a deck.
Start with one number: breakeven ROAS
Before I open a Google Ads account, I calculate breakeven ROAS. It is 1 divided by your gross margin. With a 40% margin, breakeven ROAS is 2.5. Every number in the account means nothing until it is compared to this one.
| Gross margin | Breakeven ROAS | Typical for |
|---|---|---|
| 30% | 3.33 | most accounts I inherit |
| 40% | 2.50 | typical DTC |
| 55% | 1.82 | high-margin brands |
If the dashboard shows a 3.2 ROAS and your margin is 30%, every sale loses money. The dashboard looks fine. The bank account does not agree.
Why most Google Ads reports optimize the wrong number
ROAS is a ratio. Profit is a number. Almost every agency reports ROAS. Very few can tell you your contribution margin last month or whether Google spend made money after product costs.
| What usually gets optimized | What actually moves profit |
|---|---|
| Platform ROAS (Google's number, not yours) | Contribution margin per channel, weekly |
| CTR and Quality Score (signals, not outcomes) | Breakeven ROAS vs actual ROAS, per campaign |
| Impression share at the wrong phase | Non-brand ROAS separated from brand |
| Spending the full budget | Feed quality before touching bids |
| Blended ROAS including branded search | How much PMax credit was going to convert anyway |
This is not about agencies being bad people. It is structural. Someone running 30 accounts cannot build margin-first thinking into 30 different businesses. That is why I keep a small number of accounts and run each one myself.
What this looks like in a real account
A DTC brand spending €80k a month on Google. Blended ROAS around 3.2. The account looked healthy on every dashboard, but profit swung from +18% one month to −22% the next, and they were about to raise spend by 40%. Three things were happening:
1. Branded search was counting itself
31% of “Google sales” came from people who had already typed the brand name. That is interception, not acquisition. Strip it out and ROAS goes from 4.1 to 1.7 on non-brand.
2. PMax was competing with Shopping
Performance Max and Standard Shopping bid on the same people, and each campaign took credit for the win. The reports were not lying. They just did not say which hand took the other hand's lunch money.
3. Smart Bidding was finding the worst customers
Bidding chased conversion value, so it found the cheapest €60 sale: discount-code hunters who returned half the order. They were paying Google to find their worst customers.
Results 90 days after the structural changes. Under NDA: the numbers are real, the name is not.
Five rules every decision goes through
These are constraints, not values. Every structural change and budget decision is checked against them before it ships. When a client asks for something that breaks one, I explain why and decline.
- Breakeven ROAS before anything else. The floor is calculated before any campaign work.
- Signal quality before optimization. Optimizing on bad data produces bad outcomes faster. Tracking is verified before any bid decision.
- Business logic causes campaign symptoms. Diagnosis starts at business metrics and works down to settings.
- Structure before scale. More budget on a broken structure only loses money faster.
- Scalpel, not sledgehammer. One change, one week, then we look. Changing bid strategy, budget and structure in the same week makes diagnosis impossible.
The first 30 days, step by step
Every new account follows the same sequence. Each phase ends with a deliverable, and the next one does not start until it is confirmed.
| When | Step | What comes out |
|---|---|---|
| Day 1 | Business briefing: margin, AOV, spend history, tracking state, who ran the account | Breakeven ROAS, shared on day one |
| Days 3 to 5 | Tracking audit: primary conversion, server-side status, GA4 vs Google Ads gap | A signal you can trust |
| Days 5 to 10 | Account audit from business metrics down: spend split, brand isolation, bidding, search terms, PMax, feed | A map of where money leaks |
| Days 8 to 12 | Feed and Merchant Center audit: titles, GTINs, price accuracy, images, disapprovals | Feed fixed before campaigns change |
At the end you get a written diagnosis: the structural problem, fixes in priority order, and what to expect at 30, 60 and 90 days. A document, not a pitch deck. See what it looks like.
Account structure by spend phase
There is no universal campaign structure. The right one depends on conversion volume, signal quality and spend. Most accounts I inherit run a phase 1 structure at phase 3 spend, and that is why they have hit a ceiling.
Phase 1: under €10k a month
| Campaign | Budget | Bidding |
|---|---|---|
| Brand Search | 10% | Manual CPC or target impression share |
| Brand Shopping | 10% | Maximize clicks |
| Non-brand Search | 30% | Maximize clicks with a CPC cap |
| Non-brand Shopping | 50% | Maximize clicks |
Target ROAS needs history, and at this phase there is none. Maximize clicks is not the cautious choice, it is the only correct one. No PMax yet.
Phase 2: €10k to €30k a month
| Campaign | Budget | Bidding |
|---|---|---|
| Brand Search | 10% | Manual CPC or target impression share |
| Brand Shopping | 10% | Target ROAS |
| Non-brand Search | 25% | Target ROAS |
| Non-brand Shopping | 25% | Target ROAS |
| PMax, parallel test | 30% | Max conversion value or target ROAS |
PMax is added next to Shopping, not instead of it, with exclusions from day one. Standard Shopping keeps winning non-brand intent on its own.
Phase 3 and up: €30k+ a month
Eight campaigns: brand Search and Shopping, non-brand Search, PMax split into heroes (about 40%), zombies (about 20%) and villains (about 5%), Demand Gen and advertorial Search. YouTube top of funnel joins at €100k+.
The hero, zombie and villain split forces a decision on every product group. No product keeps budget just because its campaign is running.
Tracking comes before bidding
Bad tracking produces bad signals, bad signals produce bad bidding, and bad bidding wastes budget. Browser-only tracking loses 20 to 40 percent of conversions depending on the audience. This is the foundation, not a one-time setup task.
| Layer | Tool | Role |
|---|---|---|
| 1 | Google Tag Manager | Every tag goes through GTM. Nothing is pasted straight onto the site. |
| 2 | GA4 | Diagnostics, not the bidding signal. A gap above 20% vs Google Ads means the signal is degraded. |
| 3 | Server-side tracking | ProfitMetrics when margin data is shared (sends a profit signal), TrackBee when it is not. |
| 4 | MER via Triple Whale | Total revenue ÷ total spend. If it diverges from platform ROAS by more than 30%, something is overclaiming. |
Merchant Center feed is infrastructure
A disapproved product cannot serve. A weak title cannot compete. I have seen a manager spend three weeks tuning target ROAS while the feed had 400 disapproved products. Feed work happens before campaign work, every time.
- Titles are where Google matches searches. Before: Collagen Powder Vanilla 300g. After: Hydrolyzed Collagen Peptides Powder | Skin Hair Nails Joint Support | Type I & III | Vanilla 300g | Brand.
- GTIN and MPN. Missing identifiers quietly cost impressions.
- Price accuracy. Feed price must match the site, synced automatically from the source of truth. A mismatch leads to suspension.
- Images. Text on the main image means disapproval. Clean product shots only.
- Tools: Simprosys for most Shopify stores, Channable for large or multi-market catalogs.
Bid strategy in the right order
The most common Google Ads mistake is switching to target ROAS before there is enough data. If the target is built on 8 conversions over 6 weeks, the algorithm is optimizing toward noise.
| Phase | Strategy | When |
|---|---|---|
| 1 | Maximize clicks | Under 30 purchases a month. It collects the signal everything else needs. |
| 2 | Target ROAS | Enough purchases and clean tracking. Set it 10 to 15% above current ROAS, evaluate after two weeks minimum. |
| 3 to 4 | POAS (profit on ad spend) | Margin data feeds the algorithm, so it looks for customers who bring profit. |
Maximize conversions is used only where CPA is the goal, never as a default and never on a new campaign without history.
Reporting: three sentences every Friday
The weekly report exists to make the next decision obvious, not to show effort. Every Friday you get three sentences:
Why: Impression share on the top non-brand themes fell from 34% to 12% after a theme update hurt landing page quality. Maximize clicks was spending on weak auctions.
What to expect: CPA up 8 to 12% short term, volume down 5%, profit per session up. If profit per session does not move by next Friday, the hypothesis is wrong and we revert.
Forwardable to your CFO, readable in 90 seconds. No 40-slide decks, no dashboards without context, no “we're monitoring the situation”.
Who this works for
Vaclav Hanousek · Growth Manager, DTC health and wellness
If this is how you think about your business, let's talk
30 minutes, your account on screen. I tell you what I see, and if it makes sense to work together we talk numbers then.