By now we all know that Twitter’s employee base has been absolutely decimated, going from 7500 in 2021 to somewhere between 1300 and 2300 today. Datacenters have been closed, projects have been shelved, and many former employees are speaking out. And yet that blue bird is still there, and the news headlines of “so and so posted on Twitter” still continue unabated. What happened? Shouldn’t Twitter have melted down by now?
Or… did Twitter really have that much cruft that you can cut off 70% of the entire organization and have no severe side effects? Did Musk brilliantly shave off a massive cost savings by laying off all those people who weren’t doing anything?
The cancelled projects
Corporate projects are a dime a dozen, and not all of them end up seeing the light of day. The success rate of projects in corporate environments is lower than most would expect. Many factors can contribute to a project’s failure, including poor planning, lack of funding, or a change in business priorities. Twitter is no exception to this rule, and it’s likely that many of their projects never made it off the ground.
The people laid off on such projects would certainly lose their jobs, but the company won’t miss a beat. The cancelled projects also will likely never see the light of day, but again this is not truly going to be visible to us from the outside, at least not initially.
Surely, this couldn’t be that many people involved? Yes, innovation is important, but it wasn’t two thirds of the company trying new things?
Well, maybe yes, maybe no.
Like all software companies, Twitter has undoubtedly accrued a large amount of tech debt. It could be through code that was written long ago, using an older data model, or design. It could be through the usage of open source software, which, while supported a long time, isn’t supported forever and will sooner or later become obsolete, triggering a painful migration. Often, a company will end up on woefully out of date software internally, with a heavy refactor or even rewrite process to bring the code up to date.
Punting on these types of projects has no immediate cost. Things will continue to work on the day-to-day without them. Yes, there will be consequences; it might be slower to develop new features. Things might break. Or, eventually, someone will find a bug in the old software and there is a security vulnerability. But, at least in the short term, everything is peachy.
The future sales
Business to business sales are a funny thing, compared to business to consumer. When a consumer buys a product now, we might pay $500 to a company. We instantly lose that money, and the company instantly notches $500 in revenue. Business to business sales work quite differently.
In a B2B sale, it is very typical to sign longer term commitments, such as multi-year agreements. For companies, this makes forecasting much easier; they can predict how much money they’ll earn, and how much they’ll spend. A lot of these agreements are done with elbow grease, and sales teams, and while it sounds sleazy… it is.
To take one example, in 2022, Microsoft booked $110 billion in deferred revenue. For comparison, Microsoft’s total 2022 revenue was $198 billion. Basically, they knew, ahead of time, that they would earn more than half of their revenue, guaranteed, thanks to sales people that had chalked up multi-year ahead-of-time commitments.
Twitter most likely had sales teams dedicated to cozying up with large customers and coaxing large ad buys over multiple years. The sales teams can be let go, but prepaid ad buys will remain until the money is spent or by some negotiation refunded. Until that happens, Twitter legally cannot count the money as earned.
Elon Musk is famously hostile to PR departments, and it’s likely almost, if not everyone, from that department has been let go. This would have no immediate impact on the day-to-day site operations. HR would likely slim down too, given a significant reduction in HR support needs. Similar impacts would likely be felt in office support staff, organization of events, IT support, and more, all of whom would have significantly fewer people to support. Some of that is recursive too — having fewer people in HR means less IT support needed, and fewer IT support engineers means less demand on HR.
For a large software company like Twitter, it’s not unreasonable to guess that at least 50% of the total headcount would have been supporting roles to the engineering organization. I am making guesses here because no reporting seems confirmed, and Elon Musk is contradicting almost all reporting.
Guessing the engineering bloodbath
Nevertheless, if we pull on the confirmed numbers, what we get out is that Twitter started off at around 7500 employees. If we use a range of 30-50% as engineers, that gives us a range of 2250-3750 engineers before the bloodbath.
The most pessimistic number I could find on estimated Twitter engineers today is 500 out of a total of 1000 employees. This is seemingly confirmed in a Twitter Spaces interview, and that some of the people let go were done so in error. In any event, we can make a few round guesses.
- As a smaller company, Twitter’s engineer-to-support ratio is likely going up, both as a result of less organization around it, and Musk’s penchant for completely axing some departments.
- We can reasonably assume that cuts have been harsher on general support, e.g. content moderation, than on core operations
All in all, I would probably hew closer to guessing that around 2500 of Twitter’s 7500 employees were in the engineering organization. This gives a bloodbath of around 80% of the total headcount. Not great, but it is most definitely enough to keep the site going. Quite likely a lot of secondary Twitter concerns have been getting the short end of the stick; a quick check on Twitter’s Open Source repositories shows a declining trend in many of them, unsurprising.
The diet of Twitter is severe and significant. While so, it is not enough to actually make the site crash, and an engineering organisation of 500 people is more than enough to keep it running. The big question going forward, however, is if it is enough to keep the site growing.