It's based on some arbitrary thing I did some time ago to determine how large a company should be (6-8 reports per manager and N layers of management; 1 layer = 1 + 6 + 6 * 6 = 43, 2 layers = 1 + 6 + 6 * 6 + 6 * 6 * 6 = 259, 3 layers = 1 + 6 + 6 * 6 + 6 * 6 * 6 + 6 * 6 * 6 * 6 = 1555, 4 layers = 9331, etc).
As Twitter seems to be twice as large as it needs to be (3898 vs. 1949) and as it's obvious more people will need to be hired after any cuts, I arbitrarily chose 1555 and 259/585 (same number of layers, but 8 reports, rather than 6) as the numbers since they add up to about 2000.
Was the arbitrary thing you did somewhat scientific (ie,from reading papers, etc about ideal company size)?
Asking just out of curiosity, I don't think your comment is going to be too popular. I care about this a little bit because of recent cuts I saw at Yahoo.
Twitter being twice as large as it needs to be isn't arbitrary. That's easily seen from all information available. It may not be necessary to have a round of layoffs, but it's at least clear that they've hired too many people. Rather than going the route of layoffs, they could stop hiring more people, clear out the idiot management team that's driving them into the ground, then wait until they've caught up before hiring again.
The number of employees being set to 1555 is what's arbitrary. The actual employee counts (1, 7, 43, 259, 1555, 9331, etc) aren't as arbitrary, as that's what results when one assumes 6-8 reports per manager, then calculates backward to determine the overall company size per added layer of management. That is, if you want 2 layers of management, and you want each manager to have 8 reports, then you'll have an overall size of (1 CEO + 8 executives + 8 * 8 managers + 8 * 8 * 8 individual contributors) 585.
The idea of having no more than 6-8 reports per manager is somewhat based on science (number of things one can hold in short-term memory lining up with the number of people one can effectively manage at a time), but meh.
Also, the less management layers that exist, the better. And as each manager is really only able to handle 6-8 people, once you increase employee counts past a certain amount, another layer of management becomes necessary. This could become more relaxed with better software and increased efficiency (a manager could have 21 reports, for example, if they don't have to heavily/directly manage them all), but I kept going with the 6-8 number, as it gave me a good sense of things.
Another (arbitrary?) reason those numbers seemed good to me was from assessing insertion sort vs. quicksort. That is, if you compare the two, then they are equal when there are 9 items, but insertion sort is a better choice when there's less than 9 items. It was somewhat random, but as I noticed insertion sort is how the mind naturally sorts (when it learns to be more efficient), it cast another vote for using 6-8. When there's more than 8 or 9 things in mind, one's natural tendency is to then try to break things up into chunks (like quicksort), then run the equivalent of insertion sort on each chunk, then merge everything back together.
As Twitter seems to be twice as large as it needs to be (3898 vs. 1949) and as it's obvious more people will need to be hired after any cuts, I arbitrarily chose 1555 and 259/585 (same number of layers, but 8 reports, rather than 6) as the numbers since they add up to about 2000.