I really appreciated the point about the 'effectiveness' of any data analysis. If it doesn't provide a service, it's not actually effective.
To supplement the article - I'd love to see an in depth comparison of what it takes to 'deploy' python analysis into services vs a similar problem in go.