Google Ads without burning the budget

Search intent is the entire game

Unlike social, search advertising starts with someone typing. That query is the most honest data in marketing: the customer's own words, at the moment they have a need. Your job is not to create demand. It is to be the answer.

Queries sort into four intents, and confusing them is where budget goes to die.

• Informational. "What is a CRM", "how does attribution work". The person is learning. If you sell software, you are paying for someone else's research. Rank for it organically instead.
• Navigational. "Marktstrategie login", "Meta ads manager". The person wants a specific destination. If it is your brand, own it. If it is someone else's, expect a poor return.
• Commercial investigation. "best crm for small business", "hubspot vs pipedrive". The person is comparing. Valuable, and hard, because you must win on merit rather than on being first.
• Transactional. "book a crm demo", "emergency plumber Munich". The person is ready. This is where search advertising earns its reputation.

The practical rule: your ad should answer the query the person actually typed. If the query is a question and your ad is a price, you will pay for the click and lose the visitor. If you cannot write an ad that directly answers a keyword, that keyword does not belong in your account.

Match types, and what they actually do now

Match types control how close a search has to be to your keyword before your ad is eligible.

• Exact match, in square brackets. The tightest option — but not literal. It still matches close variants, including misspellings, singular and plural forms, and reorderings that share the same meaning.
• Phrase match, in quotation marks. Your keyword's meaning must be present in the query. Google reads meaning, not just word order, so phrase match reaches further than the name suggests.
• Broad match. No syntax. The system matches your keyword to whatever it judges to be related, and that judgement can be generous. Broad match is not a synonym for "more traffic"; it is a synonym for "less control".

Broad match is not inherently bad. With strong conversion data, a target cost per acquisition and a disciplined negative list, it can find valuable queries you would never have thought to add. Without those three things, it spends your budget discovering that people search for things you do not sell.

Negative keyword hygiene is the real optimisation

The highest-return activity in most accounts is reading the search terms report and adding negatives. It is unglamorous, it never ends, and it typically saves more than any bid adjustment you will ever make.

Run it on a schedule: weekly for a new account, monthly once it settles. Read the queries that actually triggered your ads, and exclude anything that cannot become a customer. The recurring categories are consistent across industries.

• Jobs and careers. Job, salary, career, vacancy, intern, apprenticeship.
• Free and cheap. Free, cheap, DIY, template, download, tutorial.
• Learning intent. Course, certification, exam, training, definition, meaning, example.
• Wrong industry or use case. Whatever your product is commonly mistaken for.
• Irrelevant geography. Countries, regions and cities you cannot serve.
• Competitor brand terms, if you have decided not to compete there. That is a strategic call, not a default.

Build shared negative lists at account level so the same exclusions apply everywhere, and add campaign-level negatives for anything specific. Do not overdo it: a list so aggressive that legitimate queries are blocked is a self-inflicted traffic problem, and the symptom — impression share lost to targeting — is easy to misread as low demand.

Why broad match plus a broad budget burns money

The mechanism is straightforward. Broad match hands the system a wide mandate. A large daily budget removes the pressure to be selective. And a bid strategy that aims to spend the budget — as "maximise conversions" with no target effectively does, and as "maximise clicks" explicitly does — will find somewhere to put the money, whether or not that somewhere converts. The algorithm is not failing. It is doing what it was told: spend, and find traffic resembling your existing conversions. If your conversion data is thin or wrong, it is learning from a bad teacher.

The fix is to match the strategy to the data you have. Broad match belongs with reliable conversion tracking, a meaningful target and an actively maintained negative list — nothing less. Without those, phrase and exact match with a capped budget will outperform it, because they let you be wrong about fewer things at once.

Landing page relevance and Quality Score

Quality Score is a diagnostic, not a billing formula. Google reports it as a one-to-ten figure built from three components: expected click-through rate, ad relevance, and landing page experience. It is measured against other advertisers in the same auction, which means it is relative and it moves when your competitors change, not only when you do.

What it is useful for is telling you which of three things is broken. Low expected CTR means your ad does not look like the best answer to that query. Low ad relevance means the keyword, ad group and ad text are not talking about the same thing. Low landing page experience means the page after the click does not deliver what the ad promised, or loads slowly, or is unusable on a phone.

The test is simple: follow the scent. The query used the word "emergency". The ad said "24/7 emergency callout". The page must say "24/7 emergency callout" in the headline, in the same language, above the fold. If the visitor has to scroll to confirm they are in the right place, some will not bother. Send traffic to a dedicated page per intent rather than your homepage, and give each page one job.

Page speed and mobile experience are part of that assessment, and part of whether the visitor stays. A page that takes several seconds to become usable loses a meaningful share of clicks before the content is read.

Bid strategies: match the tool to the data

Bid strategies fall into two families. Manual ones — manual CPC and its enhanced variant — let you set bids and learn what a click is worth. Automated ones — maximise clicks, maximise conversions, target cost per acquisition, target return on ad spend — let the system decide, guided by your conversion data and your target.

The order that works:

• Start with manual or enhanced CPC while collecting data, so you can see what different queries and times cost.
• Move to maximise conversions without a target once tracking is reliable and volume is consistent. This spends efficiently against the conversion you chose — which is why choosing the right conversion matters more than the strategy.
• Move to a target — cost per acquisition or return on ad spend — once you have enough conversions to learn from. The commonly cited starting point is in the region of fifteen to thirty conversions in a thirty-day window, and it varies by account. Below that, a tight target starves delivery.

Two rules. Do not switch strategies often — every change restarts a learning period, and an account that changes weekly never leaves one. And never run "maximise clicks" with an uncapped budget on a campaign you care about, because its objective is clicks, and clicks are cheap in all the wrong places.

The settings that quietly waste budget

Search partners

A search campaign can serve on partner sites beyond Google search, switched on by a default most people never look at. The placements are not visible to you, the traffic behaves differently, and quality varies enormously. Measure it as a separate segment before deciding. Many accounts find it dilutes performance; some find it profitable. What is indefensible is paying for it without ever having looked.

Display expansion

Search campaigns can be expanded onto the Display Network, again by default. This changes the nature of the campaign entirely — from answering a query to interrupting a page. If you want display, run display, with display creative and display measurement. Do not let it hide inside a search campaign where its spend is blended into a cost per acquisition that looks acceptable.

Broad location targeting

The standard location option targets people in, or who have shown interest in, your chosen area — so an ad for a Munich service can be served to someone in another country reading about Munich. For a local business, that is money spent on people who cannot buy. Set targeting to presence only, and check the geographic report for where your clicks actually came from.

Auto-applied recommendations

Google will offer to make changes on your behalf: adding keywords, broadening match types, adjusting bids, pausing "underperforming" ads. Some suggestions are good and worth accepting deliberately. Automatic application is not, because it puts changes into your account that no human reviewed and that you will not find in a report. Turn auto-apply off, and treat the recommendations page as a source of ideas rather than a to-do list.

Where most people get this wrong

The most common mistake is running search ads as if they were a demand generation channel. They are not. You are capturing people who are already looking, so the ceiling is set by how many people search, and no amount of budget creates more of them. When a campaign is limited by search volume, spending more does not help — it raises your cost per click and buys the same people more expensively. The answer is a different channel, or a broader set of intents, not a bigger budget.

The second mistake is never reading the search terms report. The keywords you bid on and the queries you actually paid for are different lists, and the gap between them is where the waste lives. An account that has not been reviewed in six months is almost always paying for a meaningful share of queries that could never convert.

The third is blaming Quality Score for high costs while ignoring the landing page. It is easier to tweak ad text than to rebuild a page, so ad text gets tweaked. But if the page does not continue the promise of the ad, the click was wasted regardless of the score, and no ad copy fixes that.

The fourth is measuring everything together. Brand searches convert at a different rate to non-brand, and blending them produces a number that describes neither. Split brand from non-brand in reporting, always. Otherwise a strong brand campaign masks a non-brand campaign that is losing money, and you scale the wrong thing.

The short version

• Match the ad to the intent of the query. If you cannot answer the search directly, do not bid on it.
• Broad match only with reliable conversion data, a target, and an actively maintained negative list.
• Read the search terms report on a schedule. It saves more than any bid change.
• Fix the landing page before the ad text — message match is the highest-impact quality lever.
• Audit search partners, display expansion, location targeting and auto-apply. All four are on by default and all four spend money.