TL;DR
Fintex needed a white-label investor platform and had nothing to build on. It is sold to small and mid-size asset management firms and used by their end clients to see what they own. I was the only PM on a team of five. I prototyped it myself before the designer joined, used that prototype to open conversations with more than fifteen asset managers, and turned what came back into the roadmap. Two firms onboarded.
The problem
Asset managers in this segment cannot justify building their own client-facing portal, so their clients experience their portfolios through statements, spreadsheets and phone calls. Selling them one, white-labelled, was the bet. No asset manager will commit to a feature list before seeing something, and building something first is exactly what a team of five cannot afford to get wrong. A conventional discovery-then-spec-then-build sequence would have spent most of the runway before a single customer saw a screen.
What I owned
Everything on the product side, on a team of two developers, one designer and one QA. I ran a competitive teardown across existing investor platforms, then went cross-industry on Mobbin for the underlying interaction patterns, because making a portfolio legible at a glance is not a problem unique to finance. I built the prototype myself in Polymet and Lovable, scoped the MVP down to six surfaces (dashboard, holdings, transactions, accounts, profile and settings, plus the integrations to the pricing vendors), and ran the backlog and the UAT. On a team that size there was no separate scrum master, and no reason to want one.
The hard part
The sequencing. I built the prototype rather than writing a specification and queueing engineering time against it, and that prototype is what went to market. More than fifteen asset managers saw it. Those were sales conversations, but they were also the cheapest research I have run, and one piece of feedback changed the product outright: asset managers expected both IRR and time-weighted return, not one or the other. I had treated that as a design choice. To them it was not a choice at all, because the two numbers answer different questions and their clients ask both. The prototype bought the customer conversations, and the customer conversations bought the roadmap.
Outcome
Two asset management firms onboarded, and the build took about six weeks once the prototype had settled what we were building. The more durable outcome is the sequencing, which I now reach for by default: when the team is small enough that a wrong roadmap is unrecoverable, spend the first two weeks making something real enough to be argued with.