If you're forecasting at such a granular level 3 years in advance, your variances will be all over the place, which is not terribly useful to analyse. It won't be as per your example. It will be "I thought I'd hire one of person x, but instead ended up hiring 2 of person y, and delayed hiring z to compensate". The aggregate variance is what matters when you're doing a long term forecast.

A granular 12 month forecast is very useful for the reason you described, but we're discussing longer time horisons here.