"Getting out of the building" in metmorphical send might give you view from people / customers you can visit. This is likely to be a biased sample, and unrepresentative of the true customer.

Running A/B tests is IMO, a good proxy, to understand customers unbiased binary views on feature changes, and objectively better.