Case study · AI, data & growth
Killing an Idea with Data
My analysis showed that reselling unsold leads was worth far less than assumed, so we prioritised better regional targeting instead.
The context
Some leads were never sold to any partner. There was a strong hypothesis that reselling these leads could unlock significant value.
The challenge
I set out to understand why leads remained unsold and quantify what reselling them would realistically be worth.
How I approached it
Over approximately 18 months, I systematically analysed the causes of unsold leads. They fell into three groups: the wrong category, the wrong region, or partners having reached their limits. Only the final group could theoretically be resold, because a lead in an unsupported category or region would still have no eligible buyer later.
I then quantified the upper-bound value. On one Sunday, for solar-panel leads, 13 unsold leads represented at most €84.50 in recoverable lead costs—and only if every one of them had remained unsold because partners had reached their limits.
What it delivered
Based on the analysis, we decided not to build functionality for reselling unsold leads. Instead, we focused on root causes in our internal allocation system by increasing update frequency and targeting based on lead price rather than coverage. Not every issue received a structural fix, but the analysis helped us stop investing in an appealing idea with limited value and redirect effort toward a more promising intervention.
Looking back
I learned how important it is to separate an appealing hypothesis from the value it can realistically create. Only one of the three causes of unsold leads was theoretically addressable through resale, and the upper bound was small. I would quantify that ceiling earlier, rather than spending roughly 18 months investigating before making the investment decision.
I also learned to redirect attention from symptoms to root causes. In future roles, I will keep asking which part of a problem is actually solvable, what the best-case return is, and whether another intervention deserves priority. I value that we chose evidence over momentum and focused on improving allocation instead.