Thanks for your input Ritika! 

You’re absolutely right, an experiment should have a way to compare results between the expected and the actual, so you can actually see if an impact was made. There are actually a lot of ways of achieving this, and it’s important to remember that an experiment doesn’t need to be ‘launched’ in order to be tested.

After all, you might be able to test quite a number of things with using basic protoypes or surveys or other ways to gather information. I always recommend starting with the smallest/cheapest test possible that tests the risky elements of whatever is being put forward. Most of the time, this means you can get away with not launching something straight away, but as early experiments start giving back more info, you can get closer to experiments that involve actually launching things and trying them in Beta or behind feature flags, or whatever is appropriate at that stage.

I hope this helps! Happy experimenting 🙂

P.S. Loved the article and the way you articulated.