Google Ads Agency Berlin: The Guide to Campaign Structure, Bidding and Data Quality

Short version: Google Ads pays off most readily where demand already exists and all you have to do is collect it. The channel gets expensive when three things are missing: reliable conversion values rather than plain counts, a clean separation between brand and non-brand searches, and enough volume for automated bidding to receive signals. Campaign type and bidding strategy only matter after that.

This article works through both cases: e-commerce, where a basket sits at the end, and lead generation, where a form sits at the end and the revenue is only decided weeks later in sales. There is a worked budget threshold below for each. All amounts exclude VAT. The ranges are experience values from the accounts we manage.

Data quality, account structure, landing page and bidding strategy are not a checklist to tick off but a chain: the best account structure is worth nothing with a landing page that does not match the search query, and the best landing page is worth nothing with the wrong signals.

Which campaign types are there, and what is each one good for?

Google Ads is not a channel but a bundle of very different inventories behind one interface. Search ads, Shopping, YouTube and Discover have little in common apart from the account.

With a tight budget, the money belongs in Search on non-brand terms with clear purchase intent and, for a shop, in Shopping. Performance Max is a scaling tool, not a starting tool. Demand Gen and YouTube overlap with Meta Ads: Google captures existing demand, Meta creates new demand.

What does Performance Max automate, and what does it take away from you?

Performance Max serves across all Google inventories at once: Search, Shopping, Display, YouTube, Discover, Gmail and Maps. You supply assets, feed, target and budget, and the algorithm decides placements, bids per auction, asset combinations and audiences.

What it takes away is resolution. The channel breakdown and search terms can now be viewed in the interface, but the performance of individual asset combinations and the attribution per audience signal stay opaque. Google keeps rebuilding these reports, so check which views your account has.

That leads to a problem: a campaign that bids on your own brand searches buys cheap clicks from people who typed your name, and books revenue you would often have had without the ad. You create transparency with four levers:

• Separate the asset groups. One group per product logic. Only then can you see which part of the range is carrying the account.
• Set brand exclusions. They keep your name out of Search and Shopping inventory. They do not apply to the remaining inventories, which is why the comparison with your brand campaign is still needed.
• Use negative keywords at campaign level to exclude irrelevant and sensitive searches.
• Review search terms and the channel breakdown every month, with one question: how much of this comes from inventory you could just as well have booked on its own?

Performance Max is useful, but only alongside a clean Search structure. As the only campaign it delivers a number without an explanation.

How do you read the search terms report and build a negative set?

The search terms report shows what people typed, not which of your keywords was triggered. It is not complete: Google only reports terms above a minimum volume. So reconcile the sum of the reported search terms against the campaign cost. The gap is your blind spot.

• Sort by cost, not by impressions. You are looking for money running out of the door.
• Tie the threshold to cost, not to clicks: to the point where a term has cost more than a conversion is worth to you. At a conversion rate of 2 per cent, 50 clicks without a conversion are not proof, only a reason to look.
• Decide by intent. A term can be expensive and still be right, when the only thing that does not fit is the landing page.

A usable negative set has three layers. Intent negatives mark out non-buyers: free, no cost, do it yourself, tutorial, jobs, salary, used, template. They belong at account level. Product negatives are the variants and models you do not stock. Structural negatives keep brand queries out of the non-brand campaign.

Two technical points cost money regularly. First, negative keywords do not apply to close variants: typos and plurals are only covered through the spellings themselves. Second, a broadly excluded multi-word term only blocks when all of its words appear. The negative “red shoes” prevents “red shoes cheap” but not “red sandals”.

Which bidding strategy fits which situation?

A bidding strategy is a statement about what you can measure. If you optimise for value, you need values. If you set a target, you need enough cases to estimate it.

Target CPA and target ROAS are no longer separate strategies but target values you hand to the maximise strategies; they can only be chosen in their own right as a portfolio strategy.

You can spot too little data from four patterns. The budget is not spent despite available search volume. CPA or ROAS swing by a multiple from week to week without any change from you. Delivery comes in bursts. Learning phases never end.

Google works in rolling 30 days, not in calendar months. In our experience a target CPA works usefully from around 30 conversions in 30 days per campaign. A target ROAS needs considerably more, because the spread of the values has to be estimated as well; Google's own recommendation is higher, we work with 50. Below roughly 10 conversions, any target strategy is a matter of luck.

The answer is almost never a different strategy but more data per campaign: consolidate instead of fragmenting, portfolio strategies, an intermediate goal further up the funnel.

What data quality does the account need for automation to work?

Google's bidding algorithms are only as good as the feedback you give them. Five points decide it, in this order:

Conversion values instead of conversion counts

If every order is passed with a value of 1, no value-based strategy can tell a 30-euro basket from a 300-euro one. Dynamic values are mandatory before a target ROAS makes sense. In lead generation, forms, calls and appointments each need their own value.

Contribution margin instead of revenue

Pass the contribution margin rather than the basket value. That is revenue minus cost of goods, shipping, packaging, payment fees and the share of returns. Automation then optimises for what actually stays in the business, and high-margin products are valued above expensive loss leaders. In the accounts we see, this is rarely implemented.

One primary conversion per goal

Whatever is meant to act on the bids belongs in the account as a primary conversion action, and everything else, from newsletter sign-up to basket view, as secondary. Accounts with five primary actions optimise towards a blended value that nobody can explain.

Enhanced conversions: more attribution from first-party data

Enhanced conversions pass hashed first-party data from the confirmation page or the CRM so that more conversions get attributed. In lead generation the same mechanism works as a feedback channel: you import whether a lead turned into an order. Only then does Google optimise for customers rather than for forms.

Because personal data goes to Google in the process, this needs a legal basis, suitable consent and an entry in your privacy policy (Datenschutzerklärung). Hashing does not make the data anonymous. Whether your setup is permissible is a question for legal advice.

Consent Mode does not replace consent

Consent Mode tells Google whether consent is present and controls what is measured and what is only modelled. It is the technical transport of the consent status, not the legal basis. Basic blocks the tags until consent is given; Advanced loads them beforehand but then stores and reads nothing on the device without consent and sends only cookieless signals. What counts is not whether a tag loads but whether anything is stored on or read from the device without consent. For that you need a banner with a genuine choice. Modelled figures are estimates, not measurements.

As a cross-check, a monthly reconciliation belongs in the routine: Google conversions against shop or CRM figures. If they diverge permanently and without explanation, every optimisation built on them is guesswork. If the landing page is the bottleneck, the lever lies in conversion rate optimisation.

How much budget may go to brand search?

Brand search is the most comfortable number in the account. Anyone typing your name already knows you: click prices sit below those for generic terms, conversion rates above them. A brand campaign therefore almost always delivers a good ROAS, regardless of the work behind it. The problem is incrementality: some of those clicks would have come organically anyway.

• Brand and non-brand belong in separate campaigns. Otherwise you see a blended figure that makes non-brand performance look better than it is.
• In the accounts we manage, the brand share rarely exceeds 10 to 20 per cent of the Google budget. If it is higher, the channel is not growing, it is administering demand that already exists.
• Judge growth exclusively on non-brand performance.
• One exception: if competitors bid on your brand name, the brand campaign is defence, not growth.

This can be tested: pause the brand campaign for two to four weeks, ideally split geographically, switched off in one half of the regions and left running in the other. What gets compared is the total revenue of both areas, not the campaign revenue, and not during peak season.

From what ad budget does Google Ads make sense for a shop?

The following example is freely chosen and does not describe a real case. The threshold has two conditions: a single order has to add up, and the campaign has to produce enough orders for the automation to learn.

• Click price on generic non-brand terms: 0.80 euros excluding VAT. In the accounts we manage we see roughly 0.30 to 2.50 euros excluding VAT, depending on competition.
• Conversion rate of the landing page: 2 per cent.
• That gives 50 clicks per order, so 40 euros of ad spend per order.
• Your contribution margin per order has to be above that. At 40 euros you are at zero.

The second condition depends on the target strategy. 30 orders times 50 clicks is 1,500 clicks, and times 0.80 euros that comes to 1,200 euros excluding VAT; that is enough for a target CPA. For a target ROAS, reckon with 50 orders, so 2,000 euros excluding VAT per month.

Below roughly 1,200 to 2,000 euros excluding VAT per month for non-brand Search, a single campaign therefore stays under the data threshold. It can be done with less, but then with manual bids and a tight keyword set. If Shopping or Performance Max are added, the threshold roughly doubles, because every campaign needs signals of its own.

Worked backwards: at a contribution margin of 25 euros instead of 40, you need around a 3.2 per cent conversion rate or a click price of 0.50 euros to break even. The break-even ROAS is 1 divided by your contribution margin: at a 50 per cent margin that is ROAS 2.0, at 25 per cent it is 4.0. This formula applies to revenue-based conversion values. If you pass the contribution margin instead, break-even sits at ROAS 1.0, and your target is a contribution-margin ROAS, which is not comparable with a revenue ROAS. What outside management costs is covered in what a performance marketing agency costs.

What is different about lead generation?

Here measurement ends at the form, not at the deal. Between form and order sit qualification, proposal and close. If you optimise for forms, you get the cheapest ones. The way out is a lead value, and that is a calculation. This example is freely chosen as well.

• Contribution margin of a won order: 2,000 euros excluding VAT.
• Close rate from qualified lead to order: 15 per cent. That number is in your CRM.
• 2,000 euros times 0.15 gives a lead value of 300 euros.
• As a target CPA, take a share of that so that margin remains. At 30 per cent that is 90 euros excluding VAT per lead.
• Data threshold: 30 leads times 90 euros comes to 2,700 euros excluding VAT per month.

You do not apply this value as a flat rate across every form. Lead types are worth different amounts, and that spread is exactly what the automation needs:

The middle column gives starting points that you correct later against your real close rates. The decisive part is the feedback channel from the CRM: if the outcome flows back into the account, Google optimises for leads that became orders, otherwise for people who enjoy filling in forms.

What is different about Google Ads for providers in Berlin?

For a shop that ships across Germany, being based in Berlin is irrelevant in advertising terms. For local services it changes three things:

• Set location targeting to presence. The default setting also covers people who are merely interested in the region. That is waste. Exclude regions you do not serve.
• Steer Berlin and Brandenburg separately. Competition and travel time differ inside the S-Bahn ring, in the surrounding area and in Brandenburg. Separate campaigns make that visible.
• Treat local search intent separately. “Provider Berlin” and “near me” are different searches from the generic service term: usually a higher conversion rate, often a higher click price.

As an experience value: in Berlin more providers compete on common service terms than the German average, and click prices are higher. In the calculation above, that pushes the click price upwards, and with it the budget threshold. Further regional levers are covered in the overview of performance marketing work in the Berlin area.

What do you check first when the account is not working?

This is the order we work through in audits:

• Do the conversion figures in the account match the shop or the CRM? If not, solve that first.
• Are dynamic conversion values being passed, or does the account only count events?
• Is exactly one action per goal set as primary?
• Are brand and non-brand searches separated, and is Performance Max running with a brand exclusion?
• Does a negative list exist at account level, and when was the search terms report last reviewed?
• Does every campaign have enough conversions for its bidding strategy?
• Does the landing page answer the search query, or does the ad lead to the home page?
• Is the product feed complete? Just as relevant for Shopify shops as anywhere else.
• Is anything stored on or read from the device without consent?

If you are unsure on more than three of these points, the lever is not in the bid but in the foundations. Most of the points cost work, not budget. More on this in the article on e-commerce work.

How do we work at AdSynergy?

We are a performance marketing agency from Berlin and we work with DTC and e-commerce brands, more than 50 of them so far, with a cumulative total of more than 25 million euros in managed ad budget. On the channel side that means Meta Ads, Google Ads, TikTok Ads, conversion rate optimisation and email through Klaviyo. On Google we optimise for contribution margin. We deliberately do not promise results or timelines.

The collaboration works best from around 5,000 euros of monthly ad budget excluding VAT. That is a different number from the 1,200 to 2,000 euros above: the smaller one is the data threshold of a single campaign, the larger one is the budget from which a channel mix plus outside management adds up.

The starting point is a free, no-obligation 30-minute call. After that you receive, within 24 hours on working days, an audit of your ad accounts with concrete recommendations, following the order above. It is free and without obligation, even if no collaboration follows.

For that we need read access to the accounts or a screen share, revocable at any time. The basis is a data processing agreement under Art. 28 GDPR (Auftragsverarbeitungsvertrag, DSGVO), which we sign before the first access. We change nothing in the account without your approval. If you would rather write, you can use the contact page.

Still undecided whether to bring in outside help? Then agency, freelancer or in-house and the checklist for choosing an agency will be more use to you.

What should you take away from this article?

Google Ads scales with the search volume of your category, not with your budget. The most effective lever is the switch from conversion counts to values, and from revenue to contribution margin. Separate brand from non-brand and judge growth only on non-brand performance. Performance Max belongs alongside a Search structure. And below roughly 1,200 to 2,000 euros excluding VAT per month for non-brand Search, a campaign stays under the data threshold.

Frequently asked questions