[abeppu]

I don't know why this got down-voted so much. Maybe scaling up too fast is the biggest part of the problem.

Perhaps, putting it in a framing familiar to tech folk, if we have a new version of some important component of our application, "push it to 100% in prod before doing any data analysis because it hasn't been proven to be harmful to engagement" or whatever, probably wouldn't be acceptable to most of us. Maybe you do an A/B test and initially launch the new version to 5% and watch some stats, and if it does well, you scale up from there, incrementally expanding the audience and doing more data analysis. But that analysis depends on their being a population which you know hasn't received the new version.

The problem happens when a switch to selling stuff in plastic/plastic-lined packages is driven by parties who don't have the interest or the ability to release to only a slice of the population, collect the data or do this analysis (and to roll back when the data shows a negative impact).