It depends on the scale and latency requirements: for a startup creativity and pace of execution/person is the most important (just like Facebook in the early t

It depends on the scale and latency requirements: for a startup creativity and pace of execution/person is the most important (just like Facebook in the early times). But when you're working on a huge project with many people and many users, lots of data (while slowing down the number of new features), math gets very important for modelling the behaviour of that huge system and getting every extra $$$ out of the network effect. I myself love math, and a very bad fit for small companies, while people who just like to get dirty in throwing together features thrive there. As an example Facebook was dropping messages for a long time, and when Google started Google+, Facebook got serious in fixing bugs with consistency of its features, and started to focus on latency (using Google hires of course).