When to not use a P-value

What is a good example where the p-value is less than 0.05 but you decide not to launch a feature?

datascience example feature good launch p-value value