Failed experiments are the value of experimentation.

When we develop a new product feature, we think it is a good thing. Why else would we build it? So, without experimentation, we would launch our new feature.

If the experiment wins, then we just do what we would have done anyway. So experimentation provides no (ex-post) value.

If the experiment loses, then we do not launch the feature. Now the experiment provides value because it changes our decision.

So experiments are valuable when they lose not when they win.