One thing that has served me well in innovation is to stop treating an experiment that does not deliver the expected result as a failure.
If I learned something that brings me closer to the answer, the experiment was not a failure.
It produced data.
This has also changed the way I think about an MVP.
The goal is not to build a small version of a perfect product. The goal is to find the cheapest credible way to test the underlying assumption.
Test the hypothesis as quickly and cheaply as possible.
See what happens. Record what happened. Learn from it. Change. Test again.
And then record a little more.
When you look at the process this way, it becomes easier to let the first version be imperfect.
The purpose of the first version is not to prove that you are right.
It is to help you find out whether you are right.