It's enough to know that you'll need lots of gears and to know that you'll need high-capacity mining eventually. But there's no great harm from having a few extra or from having not quite enough to max out beaker production; just start the process, get some gears, get some beakers, and refine as needed. Without a little top-down vision, you'll end up severely overproducing some things and underinvesting in others, which is no good, but you'll never be able to build an optimal factory without first building a suboptimal one.
> Over a few months I designed an entire Factorio base without testing any of it. The goal is to create a mid-sized base capable of building a megabase. A secondary goal is to not break down or rebuild anything ever.
The challenge isn't that it's hard to figure out the ratios or math out the machines you need, it's in getting to having the production to make that base without falling asleep waiting on things. Continual bootstrapping towards that end goal is the gameplay.
Rather, I find myself trying to build a factory top down. I write a bunch of sticky notes with material requirements and calculate backwards “how many labs do I need?” “How many gears do I need to make the beakers?” “How much do I need to mine to match that hourly throughput of gears?
It’s the perfect candidate of a game to write an algorithm for. Assuming a known seed, it’s trivial to design the most efficient system based on those parameters.
And at that point it just becomes work. It’s really designed as a bottom up game, but I think about problems top down and for that reason factorio drives me nuts.