How we reduced a Series B fintech's CAC by 43% in 90 days
A full account of what we did, why we did it, and what the data actually showed. No case-study polish — the real version.
This is the unvarnished version of a project we've summarised in a case study. We've shared it here because the messy version is more useful than the clean one.
The situation
The client was a Series B fintech — consumer lending product, UK market. Their CAC was £620. Industry average for the category was £340. They knew they had a problem; they didn't know where it was.
Their attribution setup was last-click on Google Ads, with Meta and programmatic running on platform-native models. Nobody had connected the dots. Three agencies, three reporting formats, no shared signal.
Week one: the audit
We pulled 18 months of conversion data from all platforms and matched it against their CRM. The first thing we found: 31% of their "paid search conversions" were branded searches — people who already knew the product and were coming back. Paid search was taking credit for organic demand.
The second thing: their top-of-funnel content programme — blog posts, comparison guides — had zero attribution credit despite appearing in the path to conversion for 44% of new customers.
The picture that emerged was that they were spending £18,000/month defending branded search from competitors who were barely spending, while their content programme was driving nearly half of their pipeline on a budget of £3,000/month.
The reallocation
We proposed cutting branded search defence by 60% and redirecting the budget to content distribution and mid-funnel programmatic. A straightforward model — but the CFO pushed back. Understandably: you don't cut a channel that "converts well" based on a new attribution model without anxiety.
We proposed running a two-week test: cut branded search by 30%, not 60%, and track pipeline velocity in the model. If leads dried up in the model — no branded-search-assisted paths — we'd revert. They agreed.
The test showed exactly what the attribution model predicted: cutting branded search defence had no impact on pipeline. The demand was organic; the paid search was just capturing it.
The 90-day picture
With budget reallocated and attribution model corrected, the ML bid management layer could actually do its job. It was no longer optimising for a metric that was 30% noise.
By day 30, CAC was down 18%. By day 60, 31%. By day 90, 43%. Volume hadn't dropped — it had increased slightly, because the content programme was driving top-of-funnel demand that the improved mid-funnel was converting more efficiently.
The lesson
The 43% CAC reduction wasn't a media buying trick. It was the result of seeing the funnel clearly for the first time. The data was all there. It just hadn't been assembled into a coherent picture.
That's usually how it goes.