Tavo Finance sells an unsecured personal loan entirely online, competing for the same intent keywords as every bank and every other lending app in the country. Paid acquisition was the whole funnel — there was no meaningful organic or referral volume — so the cost of a lead was, in practice, the cost of the business.
Cost per lead had drifted from ₹430 to ₹640 across three quarters while approval rates stayed flat, which meant we were paying more for leads that were no better. The reporting made it hard to see why. Each ad platform claimed credit for the same lead, and nobody could say which campaigns were producing applicants who actually qualified.
I started with measurement rather than media. We moved conversion tracking server-side and fed the approval decision back as an offline conversion, so the platforms optimised toward funded loans instead of form fills. Then I rebuilt the account structure around credit bands rather than keyword themes, cut the bottom third of spend that had never produced a funded loan, and moved that budget to the two campaigns that had. The landing page was rewritten to state eligibility before the form rather than after it — which lost us volume, and gained us qualified volume.
Cost per lead fell to ₹310 and return on ad spend went from 2.1 to 4.8 over two quarters. Monthly signups grew from 3,200 to 11,400 on a budget that grew by about a third; the rest came from spending the same money on better-qualified traffic.
The account restructure got the credit internally, but the offline conversion feed is what actually did the work. Once the platforms could see which leads became loans, most of the optimisation stopped being my decision. I would do the measurement work first again, even though it is the part nobody asks for.