To be honest there are many mistakes so it's hard to say which is the biggest. One that comes to mind is data analytics. I think we implemented data analytics a lot later than we should have. Ideally you should have it as early as possible so you can make decisions based on data rather than making guesses.

Too much refinement before commercialization. No product or solution is perfect at the time that you're doing. As long as it's fitting several important pain-point or necessary, you're good to launch.

Do not afraid of failure, especially criticism from your (potential) customer; Start-up that received the most feedback is the most potential to improve to be the domain expect.