The Hydrogen Remote Monitoring work started from a falsifiable hypothesis instead of a feature list: a standalone software product for hydrogen site management would be big enough to pursue on its own. Six operator interviews broke it. The market was too small to stand alone.
A requirement tells you what to build. A hypothesis tells you what evidence would prove you wrong. "Users want a dashboard" fails that test, since nobody in an interview will disagree with it. "Operators will pay for remote-monitoring software separate from their existing platform" passes, because six conversations could contradict it, and they did.
What held up
Sway bet that group decision paralysis was a bigger problem than finding restaurants. Every persona interview landed on the same "just pick something" complaint.
What didn’t
The hydrogen concept bet that remote site management could be a standalone product. Site counts stay low through 2030, so the market couldn’t support it, and the project pivoted to extending an existing platform.
Before research starts, I write down:
- The bet, in one sentence, phrased so it can fail.
- The evidence that would prove it wrong, decided before the interviews happen.
- What we do if it fails: kill the idea, narrow it, or change the mechanism.