Most scale-ups come to Google Ads having already made the same set of mistakes. I have seen these patterns often enough — at DeepL, at other companies I have worked with as a freelance paid media consultant, and in accounts I have audited — that they feel almost inevitable. They are not. Here is what they are and why they happen.
Smart Bidding works well when it has a conversion history to learn from. It does not work well when it is optimising towards events that happen a handful of times per week. The algorithm needs volume to generalise. If you set a Target CPA campaign on day one of a new account, you are asking a model to converge on a signal that barely exists.
The standard fix is to start with a higher-funnel conversion event that fires more frequently — a sign-up rather than a purchase, a trial rather than a paid conversion — and graduate to the harder metric once the campaign has real history. Most scale-ups skip this and then blame Smart Bidding when it underdelivers.
B2B SaaS companies with long sales cycles are particularly prone to this. They set up conversion tracking on form submissions and declare that paid is working or not based on cost per lead. The problem is that most of their leads go nowhere. The conversion event they are optimising towards has little relationship to the outcome they care about.
The right approach is offline conversion imports: close a deal in your CRM, push that data back to Google with the GCLID that originated the lead. Now you are optimising towards revenue, not form fills. It requires more engineering work than a standard pixel and most companies deprioritise it. Those that do it typically see a significant improvement in the quality of leads the algorithm sends.
Broad match keywords in Google Ads have improved substantially. They can reach relevant queries that exact match would miss and they give Smart Bidding more room to find efficient conversions. They are also capable of running budget through irrelevant queries at scale if no one is monitoring.
The mistake is enabling broad match and then not building a rigorous negative keyword process. You need to be in search term reports regularly, especially in the first months. Broad match is not fire-and-forget.
A campaign structure that worked at £5,000 per month will often break at £50,000. At lower budgets you can get away with a single campaign per product area. At higher budgets you need more granular control over where money goes: separate campaigns for brand versus non-brand, for high-intent versus prospecting, for different geographic markets if CPAs differ materially by region.
The usual symptom is that you increase budget and CPA goes up sharply. Often this is not because the channel is saturated — it is because the structure cannot route budget efficiently at the new scale.
Attribution tells you which campaigns got credit for conversions. Incrementality tells you which campaigns caused them. The difference matters most for brand campaigns, where a high reported ROAS often reflects people who would have found you anyway.
Scale-ups often spend heavily on branded search and count it as performance marketing success. Sometimes this is justified — there is genuine value in owning your brand terms defensively. But it is worth running the occasional geo holdout or conversion lift test to understand what is actually incremental before treating brand CPA as signal about the health of your acquisition.
If any of this sounds familiar and you want a second pair of eyes on your account, get in touch.