There’s a pattern that repeats itself in almost every Google Ads account I review: the problem isn’t in the bid or the ad, it’s in the keyword list. And not because terms are missing, but because there are too many, they’re poorly grouped, or they respond to an intent the advertiser never stopped to analyze. This article isn’t a keyword research guide (there’s already plenty of material on that). Here we get to what’s expensive: the specific mistakes that inflate cost per conversion and how to detect them before the budget is gone.
Why most keyword audits start with the search terms report
Before touching anything, the first thing I look at is the search terms report from the last 60-90 days. Not the keywords you declared, but what people actually typed. According to Google’s official documentation, this report shows the actual terms that triggered your ads, and that’s where the gap appears between what you think you’re buying and what you’re buying.
A real example I remember clearly: a physiotherapy equipment store had “folding table” as a keyword in broad match. It was paying for searches for “folding table for baby”, “folding camping table” and even “cheap folding bed”. None had professional purchase intent. The CTR was decent, so in the dashboard it looked like a healthy keyword; the problem was in the cost per conversion, three times above the account average. Adding negatives for baby, camping and bed lowered the spend without touching the bid. That kind of adjustment you don’t see in the keywords dashboard: you see it in the search terms.
Mistake 1: confusing broad match with “I’ll filter it later”
Broad match has changed a lot. It’s no longer just synonyms: Google applies what it calls “semantic similarity expansion,” and that includes reformulations that sometimes drift quite far from the original term. The result is that a broad keyword can be triggered by searches that only share the topic, not the intent.
How to avoid it without giving up reach
- Start with phrase or exact match for the keywords that sustain the business and leave broad only for discovery, with a separate budget and a specific ad group.
- Review search terms every week during the first month of a new campaign. After that, every two weeks. It’s not optional.
- Use the “matched keywords” column in the search terms report to see which keyword triggered each search. If a generic keyword triggers terms that don’t fit, lower the bid or pause it.
A caveat about limitations: phrase match isn’t a mathematical guarantee either. With low search volumes, Google can ignore some words in the phrase and trigger the ad anyway. There’s no such thing as perfect match; there’s match that gets monitored.
Mistake 2: grouping keywords by lexical similarity instead of by intent
This is the structural mistake that costs the most in the medium term. Putting “home physiotherapy price” and “how to become a home physiotherapist” in the same ad group because they share words is an intent mistake, not a semantic one. The first is looking to hire a service; the second is looking for professional information. The ad and the landing page can’t be the same.
The practical consequence is twofold: CTR drops because the ad doesn’t meet the expectation, and the quality score suffers, which makes the bid more expensive. The solution isn’t to create a hundred ad groups with one keyword each (that fragments the history and complicates management), but to group by what the user wants to do when they click:
- Transactional intent: “buy,” “hire,” “price,” “near me.”
- Informational intent: “how,” “what is,” “guide,” “differences.”
- Comparative intent: “best,” “vs,” “alternatives.”
- Brand intent: terms with the competitor’s name or your own.
If you work with tight budgets, informational keywords in Search are usually a drain if the landing page isn’t prepared to capture a lead. That’s where it’s worth asking whether that keyword should be in Search or shouldn’t be there at all.
Mistake 3: negatives that get added once and forgotten
Negative keywords aren’t a static list. They change with the seasons, with trends and with search behavior itself. A negative list from a year ago protects you from last year’s terms.
What to review and how often
A reasonable routine: review the shared account-level negative list every quarter and the campaign-specific ones every month. Pay attention to three categories that tend to slip through:
- Employment terms: “job,” “employment,” “salary,” “exams.” If you sell training, they might interest you; if you sell a service, almost never.
- Free terms: “free,” “download,” “template.” It depends on the business model, but in professional services they’re usually noise.
- Unserved geographic terms: if you only operate in three provinces, any other one is spend with no return.
A detail few take advantage of: negative lists can be applied at account and campaign level, and it’s worth having one shared master list plus local lists per campaign. Duplicating negatives in every campaign is work that’s lost as soon as someone restructures the account.
Mistake 4: ignoring the match between keyword, ad and landing page
Quality score isn’t a mystery: Google scores ad relevance, landing page experience and expected CTR. If the keyword is “van rental Madrid” and the landing page is the generic homepage of a car rental company, the user clicks, doesn’t find what they were looking for and goes back. That bounce translates into a worse quality score and, over time, more expensive bids for the same position.
The practical rule: each ad group should have a landing page that literally responds to the intent of its keywords. You don’t need a new page per keyword, but you do need one per intent. If you have ten ad groups and they all point to the homepage, you have a relevance problem, not a keyword problem.
Mistake 5: measuring a keyword’s success by CTR instead of cost per conversion
A high CTR can hide a keyword that attracts curious clicks but not buyers. In accounts with contact forms, I’ve seen keywords with an 8% CTR and zero conversions in two months, while others with a 2% CTR sustained half the leads. The metric that decides whether a keyword stays or goes is cost per conversion (or return), not the click percentage.
That said, there’s an important nuance: in brand campaigns and cold traffic acquisition campaigns, CTR is a useful signal of ad relevance, though not of profitability. Don’t mix performance metrics with relevance metrics in the same decision.
Mistake 6: not separating brand from non-brand in the structure
Mixing brand searches with generic searches in the same ad group distorts everything. Brand usually has high CTR and low cost, and it contaminates the reading of the generic ones. Moreover, if someone searches your name and a competitor appears, bidding on your own brand is a strategic decision different from capturing new traffic.
The minimum recommended separation: brand campaign, generic transactional campaign, competitor campaign (if you decide to run one) and discovery campaign. Each with its own budget and cross-negatives (the generic one should negativize brand terms to avoid cannibalizing).
Mistake 7: copying keyword lists from tools without filtering by business
Suggestion tools generate enormous lists. The problem isn’t the quantity, it’s that many include terms that have nothing to do with your business model or your service capacity. A dental clinic with two locations shouldn’t bid on terms for cities where it has no presence, no matter how much volume the tool shows.
Before exporting a list, filter by three criteria: can I serve that search?, do I have a page that answers it?, does the margin justify the estimated cost per click? If the answer to any is no, the keyword doesn’t go in.
A case that illustrates the cost of not reviewing intent
A client in the online training sector had a campaign with the keyword “Excel course” in broad match. Spend was high and conversions scarce. On reviewing the search terms, a significant portion of the spend was going to searches for “free Excel course” and “solved Excel exercises”. The keyword was correct; the match type and the lack of negatives weren’t. Adding negatives for free, exercises and download, and switching to phrase match, reduced spend on irrelevant terms without losing conversion volume. It wasn’t a brilliant optimization: it was a review that should have been done in the first week.
What doesn’t work (even though it’s repeated a lot)
Bidding higher to compensate for a keyword with a poor quality score doesn’t fix the problem, it makes it more expensive. Nor does adding dozens of generic negatives without reviewing the actual terms work: you can block searches that were converting. And no, there’s no “ideal number of keywords per ad group” formula. It depends on intent, volume and the ability to maintain specific landing pages. A group with a single keyword can be correct if that keyword has high volume and a very defined intent; a group with fifteen can be correct if they all share intent and landing page.
A minimum routine to avoid repeating these mistakes
- Weekly review of the search terms report during the first month of each campaign; after that, every two weeks.
- Monthly audit of the shared negative list and the local ones.
- Quarterly check that each ad group has a landing page that responds to its intent.
- Clear separation of brand, generic and competitor from day one.
- Decision to pause or keep a keyword based on cost per conversion, not isolated CTR.
None of these practices guarantees results: the market, competition and seasonality all play a role. What they do is prevent the budget from going to searches that were never going to convert.