Validation goes wrong in predictable ways. The first mistake is asking leading questions — "Would you use an app that does X?" invites a polite yes. Ask about behavior instead: what did you do last time you faced this problem?
Second, founders talk to friends and family, who won't tell them the hard truth. Third, they confuse interest with demand; a signup is not a sale. Fourth, they build too much before testing — a landing page beats a finished product for learning.
Other traps: ignoring competitors ("we have none" usually means no market), sizing a market on wishful thinking, mistaking a feature for a business, and quitting after the first no instead of learning why.
The fix for all of them is the same: seek disconfirming evidence, not applause. Design each test so it can prove you wrong. If your idea survives honest scrutiny, you'll build with far more confidence.