Engagement is a pretty shitty metric when used in isolation though. I've worked at places where changes that increased engagement were considered successful at higher levels even though the support team knew the increased engagement was caused by increased confusion and friction in using the product.