Okay, I've really struggled to make title both short and descriptive enough, but here is a longer story.

I'm really fascinated by reading posts from dev blogs where a team facing some problem or need for optimization, took very smart trade off, based on data analysis or just doing a lot of brainstorming and putting a lot of thought into understanding the problem. Trade offs that are specific for their domain and allowed them to choose non-obvious decision to drastically improve their software design.

Examples:

- Twitter redesigned their media storage system, based on data analysis. They found that images older than 20 days has very low probability of being accessed, so it's ok to precompute thumbnails on the fly. Sort of storage/cpu tradeoff, saved them 4TB per day.

http://highscalability.com/blog/2016/4/20/how-twitter-handles-3000-images-per-second.html

- Ravelin was about to use heavy graph databases for the fraud detection purposes, but after thorough analysis of the problem they come up with the solution that only uses Union Find datastructure.

https://skillsmatter.com/skillscasts/8355-london-go-usergroup

So I'm looking for more examples like these, when understanding the data you work with lead to not very obvious/popular tradeoffs, that simplify or improve the overall software design drastically.

If you know such examples, links, talks, blog posts, or can tell your story, please share!