What to build, and why.
Product judgment is learned. Mine came the long way around. Journalism, marketing, IT, business operations, then product. Each one taught me to look for evidence in a different place, and to notice when the loudest version of a thing isn’t the truest one.
Where I find signal
Actions in the product.
Customers, frontline associates who are themselves users, and UX research arriving as findings that need conversation. What I’m looking for is the workaround. Somebody built a small habit to get around the product, which is evidence the design failed them, not that the software broke.
Impacts to the product.
Market and competitor signal, usually arriving with executive conviction already attached. It becomes a hypothesis, handed to me or built with me. My job starts after that: validate or invalidate, shape what the bet looks like in practice, work go-to-market and pricing, define the return we expect.
Expert voices on the product.
The most underused of the three. Engineers, architects, and QA hold the clearest picture of what’s newly possible, and almost nobody asks them what problems it opens up. This predates AI. AI widened the gap between what can be built and what product leaders know to ask for.
What I do with it
Relief before improvement.
Somebody is working around something today. Take that away first. Not because it’s small or easy, but because the workaround is the evidence that the design failed them. The broader fix can wait its turn.
Keep watch, then act accordingly.
Known structural weak spots get flagged and monitored, not immediately acted on. Gather evidence. Find the pattern. Assess when and how much to act.
Evidence, not elegance.
There’s always a pull toward the right answer instead of the available one. The clean model, the real fix, the thing built to last. It’s usually genuinely better, and it’s slower, and the customer waits while you build it. That’s the trade. The business gets a cheaper next time. The customer pays for it in delay.
One rule holds across all of it. How loudly a signal arrives says nothing about whether it’s right. The quiet workaround is often the true one. The loud mandate still has to be proven.