Find a real problem, confirm it is worth solving, and lock what makes the cut. The last cheap phase.
a sign you are ready: real pain, defensible, scoped
what happens inside, and how to do it
Separate the real problem from the feature request. A signal is evidence a problem exists, not the solution to it.
howGather signals from wherever they live (your own life, people you know, reviews, data), write them down, and state the problem as one specific sentence: who cannot do what, and why. Not 'they want dark mode.'
helps: clusters reviews and forums into patterns in seconds.
watch: it inflates two mentions into 'everyone wants this.' Check it against the raw source.
common trap: solutioning before you actually understand the problem.
Two questions: is the problem real enough to solve, and is solving it worth the effort?
howTalk to five to ten real people about their current workflow, not 'would you use my thing.' Listen for genuine pain over politeness. The strongest signal is someone already hacking a workaround, or willing to pay.
helps: builds the numbers and drafts your interview questions.
watch: synthetic 'users' are an echo chamber. AI cannot feel real pain; a person has to.
common trap: leading questions, and treating a survey as validation.
Choose the one thing. Saying yes to one is saying no to everything else.
howLock your kill-criteria up front: the problem, who it is for, the smallest real version, the evidence that would make you quit, the metric for success. Then scope to the bicycle, the smallest thing that does the core job well.
helps: will argue against your pick to surface blind spots, if you ask it to.
watch: it hedges and seeks consensus. The conviction has to be yours.
common trap: building five things poorly instead of one thing well.
This is one way to know Spark is done: real pain, defensible, scoped. But it is not a locked gate. You decide what this phase means for you and when you are ready to move. Read around it, think through the what and the why, then write your own definition. an hour spent well here saves weeks downstream.