The most expensive way to scale a Google Ads account is to raise the budget before you know what counts as a sale. When that happens the dashboard shows clicks, the business shows revenue, and nobody can explain the gap. The first ninety days are not paperwork: they are the installation of the instrumentation.
A conversion that means something
Before touching campaigns, we decide which action counts. “All conversions” is not an answer: a download, a click-to-call and a completed form are not worth the same. We pick one or two actions that can be tied to money (quote request, purchase, booking) and mark them as primary.
There are two common ways to measure them: the Google Ads tag or an import from GA4. Both are valid; using both for the same action and counting it twice is not. If the business closes sales over the phone or in the shop, web conversions undercount, and that is where offline conversions or a CRM join comes in.
In the European Economic Area, consent mode also has to be in place so the signals from people who decline cookies are modelled rather than lost. That is not a legal footnote: without it, part of the funnel disappears from the reports and decisions get made on half the picture.
Search terms: the list of what you are not
In the first months, the search terms report gets reviewed every week. Google hides low-volume queries, so it is not the whole picture, but what does appear is enough to build a negative list that saves real money: competitor brands, services you do not offer, cities you do not work in, job and second-hand searches.
With broad match, that list stops being an optimisation and becomes part of the structure. Broad match without negatives is a donation.
The landing page, not just the ad
An ad promising “a quote in 24 hours” and a page that takes six seconds to load and asks for seven fields is a leak. We check three things before scaling: that the page says the same thing as the ad, that the form asks for the minimum, and that the mobile experience actually works. This is where the previous two notes meet: performance and clarity matter more than the ad copy.
When you can scale
Google suggests roughly 15 conversions in 30 days for Target CPA to have signal, and 50 for Target ROAS. Those are thresholds, not guarantees: below them, the system learns from noise.
Three rules we apply:
- Do not judge smart bidding in two days. The learning period can last a couple of weeks; changing it every morning restarts the process.
- Raise budget in steps. Increments around 20% let the system adapt without breaking CPA.
- Read the right window. The default window is 30 days from the click, so last week’s campaign has not finished counting yet.
What we look at in week twelve
Not cost per click. We look at cost per qualified enquiry, which terms bring customers, and which campaigns produce sales the CRM can close. If that chain does not exist, the next step is not more budget: it is fixing the chain.