A roofing company in Texas came to us with a Google Ads account that looked, on paper, like it was doing well. Cost per lead had dropped from $70 to $31 over six months. The previous agency had put that number at the top of every report.
The owner's problem was that his closer was spending every morning calling people who wanted a gutter cleaned, a quote for a shed, or a job three counties away. Of the 140 leads that month, nine turned into a site visit and two into a signed roof. The real cost per customer was over $2,000, and nobody had been tracking it.
This pattern shows up in almost every high-ticket account we audit: law firms, dental and aesthetic clinics, HVAC and solar, B2B software, financial advisers. The account is doing precisely what it was told to do. It was just told the wrong thing.
How the account learned to find bad leads
Google's bidding does not know what a good lead is. It knows what a conversion is, because you told it. If the conversion action is "form submitted" or "call lasting 30 seconds", Smart Bidding will go and find the cheapest possible people who submit forms and make 30-second calls.
Those people are disproportionately price shoppers, people outside your service area, people with the wrong problem, and students doing research. They are cheap to acquire because nobody else wants them. Over months, the algorithm gets steadily better at finding more of them, cost per lead falls, and the report looks better while the business gets worse.
The previous agency on the roofing account had also switched to broad match keywords "to let the algorithm learn", which meant the account was bidding on "roof" in every context Google could imagine. Gutters, sheds, repairs, DIY questions. All of them produced cheap form fills.
Fixing the conversion first
Before touching a keyword, we change what the account counts as success. There are three tiers, and the goal is to get the bidding to optimise on the deepest one you can measure reliably.
Tier 1: qualified lead. Not every form fill, only the ones that pass a filter. For the roofer that meant: inside the service area, residential, full replacement or major repair. We added two questions to the form and a service-area check, and only fired the conversion when the answers qualified. Lead volume halved overnight. Nobody minded.
Tier 2: booked appointment. When the CRM or booking tool can send a signal back to Google, whether through offline conversion import, a webhook, or a simple Zapier step, the account can bid for people who actually book. This is where most high-ticket accounts should live.
Tier 3: closed revenue. Importing actual sale value back into Google Ads with enhanced conversions for leads. It is more work to set up and it is the single highest-leverage change in high-ticket paid search, because the bidding can finally see the difference between a $400 repair and a $24,000 replacement.
We got the roofer to tier 2 in three weeks and tier 3 in two months. Cost per lead went back up to around $85. Cost per signed job dropped to under $600.
Then the keywords
With the conversion fixed, the keyword work becomes obvious rather than speculative.
Pull the search terms report for 90 days and sort by spend. Mark every term as one of: buying intent for your core service, adjacent service, research, or wrong entirely. In most accounts we review, 30% to 50% of spend sits in the last three buckets. Those become negatives, and the broad match experiment gets reined in to phrase and exact around the terms that actually produced booked appointments.
Then look at what is missing. High-ticket buyers search differently from bargain hunters: "roof replacement cost Austin", "emergency roof repair near me", "metal roof vs shingles", "roofing contractor reviews". Build ad groups around the intent stages, with ads and landing pages that match. A person comparing materials should not land on a "get a free quote" page with a countdown timer.
Landing pages that qualify instead of just convert
The instinct with lead generation is to remove friction: short form, big button, no questions. For high-ticket services that is backwards. You want friction that filters.
That means stating the service area on the page, stating a starting price or a typical range, asking the two or three questions that tell you whether this is a real job, and making the next step a scheduled call rather than a vague "we'll be in touch". Conversion rate will fall. The ratio of booked calls to form fills will rise sharply, and that is the number you are paying for.
What this looks like in a report
If your agency is reporting cost per lead and nothing else for a high-ticket business, ask them for three more columns: qualified leads, booked appointments and, if you can share it, closed revenue. If they cannot produce those, the account is not being optimised for your business. It is being optimised for the report. That question, and eight others worth asking before you sign anything, are in the agency screening list — and if you are still deciding whether to hire an agency at all, the honest comparison with a freelancer or an in-house hire comes first.
If the buyer is definable by job title and the deal is large enough, LinkedIn is often cheaper per customer despite costing far more per lead — the same measurement gap decides both.
If volume is up and revenue is flat, the account is almost certainly buying the wrong people: that is a qualification problem, not a traffic problem.
Two things usually need fixing alongside the conversion action. The first is the search terms themselves, covered in why your missing negative list is the real budget leak. The second is upstream of the ad account entirely: if your brand reads like every competitor, your cost per lead climbs no matter how well the account is run.
It is also worth checking where those clicks land. A qualified searcher dropped onto a general homepage will disqualify themselves, which is why paid traffic belongs on a page built to answer it.
The free growth audit covers this for Google Ads accounts: we read the search terms, check what the conversion action actually fires on, and tell you how far your account is from bidding on real customers.