In April 2025, a Sri Lankan B2B fintech came to us with a familiar problem: their sales team was idle. Marketing was sending them leads, but 60% of them were fake emails, students, or competitors. The CEO was paying a 4-person sales team to chase ghosts.
Six months later, we'd delivered 812 verified leads — independently audited at 95% verified human — and their close rate climbed from 11% to 28%. Here's the playbook.
The audit: what we found wrong
Their existing lead-gen agency was running broad-targeting Meta campaigns. The hooks were generic ("Learn more about our platform"), the landing pages had 14 form fields, and there was no qualification step. Cost-per-lead looked great on paper at $4.20 — but cost-per-qualified-lead was over $80.
The fix wasn't more spend. It was tighter targeting and brutal qualification.
What we did differently
- Narrowed audience by job title and company size. Cut the audience from 2.4M people to 38K. CPM went up. Quality went up more.
- Replaced the form with a calendar. If you can't book a meeting, you're not a lead. Tire-kickers self-select out immediately.
- Added a $0 verification step. A short qualification quiz before the calendar. Anyone who fills it out is serious.
- Layered LinkedIn retargeting. Anyone who hit the landing page got served LinkedIn ads with social proof from their industry.
The numbers, six months in
812 leads. 95% verified human (third-party audit). 28% close rate on demos. Average deal size up 40% because we attracted senior decision-makers instead of analysts looking to learn.
The cheapest lead is the one that closes. Everything else is noise dressed up as a metric.
What we'd do differently
Honestly? Move faster on the LinkedIn layer. We waited until month 3 to add it because we wanted clean attribution data. We should have shipped it in week 2 and figured out attribution later. Clean data doesn't pay the bills.
Want a free 30-minute audit of your funnel? We'll show you where leads are leaking and exactly what we'd fix in your first 30 days. No pitch unless you ask.
