Just a day after Unity announced it would be laying off 1,800 employees as part of an ongoing “company reset”, it’s bei…
This just figures after their tvos app gets worse and worse.
Why does Twitch even need 500 employees? Surely they just need a dozen or so developers and a handful of IT people for each datacenter (assuming colo).
I guess I always assumed they had something like 100 employees.
Twitch builds and maintains the following infrastructure:
- Website + web video player
- iOS App
- Android App
- Fire TV app
- Playstation 4 app
- Xbox One app
- Chromecast app
- Apple TV app
- Real-time chat backend
- Search services
- Distributed video storage backend services (VODs)
- Content recommendation services
- Account services (streamer accounts and viewer accounts)
- Monetization services (subscriptions, bits)
- Twitch Studio + Soundtrack apps for creators
All of this has to run at Twitch scale (140 million MAUs)
And these are just the technical teams. Then add on UX designers, marketing, product and business development, not to mention Business Intelligence data scientists.
A minute after posting this, I can think of way more necessary roles. Let’s start by mentioning that all this infra needs to run 24/7 with 100% uptime so that some 30 year old can jerk off to a VOD of Amouranth at 3am without a single frame drop.
And all of this is just core product people.
Then you need HR, managers to hire, fire & promote all these people, lawyers, customer service reps, content moderators, executive assistants, and the facilities maintenance people who refill their snack closets.
Its hard to run a huge online service like Twitch.
Interesting you said exec assistants and not execs, almost like the empty suits are the least valuable addition to a company. hmmmmm…
following infrastructure
A lot of that stuff doesn’t need regular work, so maintenance should be pretty limited. I could see a team or two (~10 people) maintain most of that. Most of it really doesn’t need much, just updates to stay on top of security threats and whatnot. So like max 50 people there, at least on the development side, and that’s being generous.
Most of the work is going to be DevOps and other IT roles for maintaining databases, load balancing, etc. But again, that doesn’t need a ton of people, as in, I could see the whole company being under 500 employees easily, and that’s with a fair amount of redundancy. It’s not like they’re doing much R&D for new products, just iterating on engagement metrics.
Twitch scale
Hence the IT people.
You’re really showing that you have absolutely no idea what any of this takes. These teams take large chunks of people - their live service ops alone is probably 300 people, not to mention client application teams, teams working on new features, moderation teams, support teams, billing teams, QA teams, and then literally all the execs, management, PR, HR, finance, design, office staff, and the whole slew of other things every large company needs. The funny thing is the person already listed all of these out for you… It’s pretty clear you don’t know what this all involves and haven’t worked at a company like this at all.
I’m sure it’s over 1,000 people.
I really don’t think that’s necessary.
My company is a few thousand people, most of those are blue collar factory workers producing or delivering our core product (we design custom trucks for delivery because our product is very dangerous). Our actual corporate staff is like 200-300, and that includes a lot of people handling customer stuff (we’re B2B, so marketing, payment processing, etc is mostly manual and catered to specific clients). Our dev team is 40-50 people with four release channels (web, Android, iOS, and Windows desktop app), and is broken down into 7 distinct 5-6 dev teams that are all cross functional and can deliver all products. We do a lot of complex simulations and new feature work, so it’s not some boring product ordering service. If we were maintenance-only, we could probably cut that to less than half.
I understand that Twitch has very different problems and thus needs way more staff than us, I’m merely giving an indication that I know how projects generally scale.
Let’s look at a somewhat similar company, Valve. Valve has ~360 employees and often has ~11M concurrent users (i.e. people playing a game using their service). They probably don’t push as much data as Twitch, but they’re in the ballpark, and they build a lot more compelling products, like Steam Deck, Steam Link app, and video games. They’re a combination hardware company, game studio, and CDN. Twitch is pretty much just a CDN. I highly doubt the increased load Twitch handles to really need >5x the employees (assuming 500 employees let go is <25% of the total company).
- youre not in the same industry
- you havent actually worked in or have experience with companies like Twitch
You sure talk a lot about something you know nothing of consequence about.
Then enlighten me. What makes Twitch need way more employees than Valve?
Scale and bandwidth. Valve is not in the same industry, and you cannot use it as a comparison.
Valve isn’t operating streaming apps that require both the bandwidth control and continuous storage. That alone complicates it several fold. Valve is a store and game distribution site. The scales are wildly different.
Live Streaming and hosting videos over multiple platforms is a wee bit more complicated than just providing downloadable files.
The largest work is probably content moderation and support, but with that it is very likely, that steam support is not employed at valve, but contracted out.
Do you have experience in big tech? Because its completely normal for each of these app teams to have 3-5 developers at minimum, plus a manager, a product person and likely a QA as well. Even when not working on brand new features, these teams are all running A/B tests, working on marketing campaigns, keeping the SDKs and service frameworks up to date, responding to help requests etc. Every time there’s “new bits”, for example, its because a team of people made that.
You don’t have to believe me but I’ve personally worked in systems like this and there’s more copmlexity than you’re imagining.
Twitch actually has minimal IT - they host on AWS because they’re owned by Amazon. They pay a discounted rate but otherwise don’t maintain their own server farms or hardware.
Yes, I work in a medium-sized company with ~40 devs across 7 teams. Our company has thousands of employees total, but only about 40-50 software devs because we’re a manufacturing company (we do lots of cool stimulations though, so not a “boring” B2B that just handles forms). Each team has 5-6 devs, one QA, and there’s a design and DevOps team that everyone has access to (total team is something like 60 people). We maintain four release channels (web, Android, iOS, Windows desktop), plus an externally facing API. We don’t have a ton of customers (B2B with customers in the thousands), so we don’t need to scale much, but we do have a much faster pace of new features than many larger companies. Since we’re B2B, at have far fewer support than normal for an app of our complexity (we have two), so add another 50 or so less technical people to handle that.
We do cross-functional teams with each team being able to work on all channels. For an app like Twitch without a ton of changes, it’s not unreasonable to expect a single team to maintain multiple products. I think my org is capable of handling everything except infra for something like Twitch, so that’s why I estimate the need for developers in the dozens, not hundreds.
If infra actually is handled by AWS, there’s even less need for headcount if they can offload their CDN. Things like scaling can be automated and only need a team or two to investigate if there are problems.
so we don’t have to scale much
I think this is the part that you’re taking for granted. “Customers in the thousands” is no where near the scale of Twitch, and I can guarantee that Twitch uses more bandwidth a day than your company uses in a year (but probably multiple years).
Sure, but bandwidth doesn’t require manpower, especially if someone else is handling infra. Once you’re operating “at scale,” an increase to infra load requires only modest manpower to maintain.
My only relevant experience here is on the dev side, and from the outside, it looks like Twitch could operate fine in terms of app maintenance with my org’s dev team (like 40 people), or at least in the same neighborhood. So the rest are other IT support staff (DevOPs, DBAs, etc), which really shouldn’t be more than 100, especially with AWS managing the HW. If they’re laying off 500, that means they probably have >2k. I really don’t see how they’d need much more than 500 total, especially if we assume that Valve is comparable in terms of complexity and seems to have <500 employees.
I’d like to see a breakdown of Twitch by role, because I’m not sure what kind of roles these 500 employees would be.
Bandwidth doesn’t require extra manpower? Since when? If you have a solution to reduce manpower for massive amounts of daily bandwidth, you could change big tech forever. So please enlighten us.
Where are you getting the idea that having that much traffic, bandwidth, and daily active users only requires a skeleton crew? It’s clear to me that you’ve never worked at scale, and never have you had to deal with anything related to big tech (and I have, which is why I know that you’re incorrect). If you ever go to a bigger company, you are in for a shock. You really think a huge company like Twitch just has dead weight that they pay salary to for no reason, when their goal is ultimately to make more money?
It’s okay to be wrong. What’s not okay is continually refusing to admit that you are in the wrong when multiple people are telling you exactly what the flaw in your logic is. You clearly are not worth arguing with, because you always have to be right and can never be wrong.
Like with Twitter when Musk fired all these unproductive employees, suddenly even the backend went quickly downhill. Twitch will have a lot of social policing going on and deals with content creators and advertisers.
Sure, but if they’re firing 500 employees, they probably have >2k total. That’s a lot! How many salespeople and content moderators do they need?
Based on their recent rule changes, not that many content moderators
Content moderators would be fairly labor intense.
Content Moderation at this kind of scale is:
-
Impossible to do without both humans and algorithms
-
Always going to produce absurd, hypocritical and controversial interpretations of and changes to the TOS
-
Cause a horrific workload for those programming the algorithm, who must devise ways of screening for not actually well defined and constantly changing TOS no nos
-
Literally traumatize and cause massive mental damage to the human moderators
Facebook has these problems on an even worse scale, and still operates basically computer equipped sweatshops of hundreds and thousands of people in less economically developed parts of the world, most of whom report massive mental trauma from having to constantly review absolutely horrific content, day in, day out, for years.
100% agree, but the alternative is giving the average user moderation power and hoping they do a good job at it, or not moderate at all.
Thats /an/ alternative.
Another alternative is /social networks this large should not exist/.
There are many, many other alternatives.
Its just that social networks this large have basically destroyed the brains of people who use them, so now they can hardly imagine alternatives.
And that is /another/ argument for why they shouldnt exist, the fact that they normalize themselves they way social and cultural institutions do, but with no actual accountability the way that local and state governments at least theoretically do.
This is also an explanation of why such things are not likely to go away. In addition to being addictive at an individual level, the network effect causes peer pressure to engage more, and otherisizes those who do not and makes them social outcasts, at least amongst the relevant ages ranges for given platforms, but this has also already become more pervasive in matters of direct economic importance, with many companies not hiring, and apartments not renting if they cannot first verify your social media presence on these large platforms.
To slightly inaccurately quote Morpheus from Deus Ex:
The human being desires judgement, without this, group cohesion is impossible, and thus, civilization.
At first you (humans) worshipped Gods, then, the fame and fortune of others. Next, it will be autonomous systems of surveillance, pervasive everywhere.
Welp, turns out that real life mass scale social media networks literally are a hybrid or synthesis of the elements of the latter two mechanisms of social reverence/judgement.
-
Fair, but surely a lot of that is automated, no? You’d want a human to review it, but it’s not like you’d need people watching the streams constantly.
I’m just saying that eliminating 500 people means they have a lot more than 500 people working there, probably well over 2k. That’s way bigger than I expected.
I think you greatly underestimate how large of a platform Twitch truly is. They have over thirty million daily active users.
Probably. I only watch one streamer, and only occasionally.
That said, headcount shouldn’t need to scale much with more users. Look at Valve, which has ~360 employees and hit 33.5M active users, ~11M playing a game. Here’s some of what Valve does:
- hardware products, like Steam Deck and Valve Index
- Windows compat - Proton; granted, most of the people working on this aren’t Valve employees, but contractors Valve pays
- make games - not often, but there’s still maintenance work
- manage a CDN - not quite as much data as Twitch, but still substantial, and it’s certainly in the realm of not being a huge difference in terms of manpower to maintain
- Steam Link app - available on many of the platforms you listed
- Steam mobile app
- Steam app - Linux, Windows, macOS
So Valve has a similar-ish level of complexity with well under 500 employees. Maybe Twitch needs another 100 or so employees to manage the CDN, but surely not another 1500 or more.
but surely a lot of that is automated, no?
You know, ppl have tried to automate it in its totality. They’ve tried to make it 50/50, but it turns out there’s not much to automate. Sure you can automate copyright claims on media sources, but that’s about it. As soon as there’s any complexity to it, human review is necessary. You have to appreciate that content moderation mistakes can have a ripple effect into platform integrity and company image as well as user experience. The risks are easy to underestimate.
Oh sure. I’m just saying that a big chunk of it can be automated, so you’re left with manual review of clips that either users or bots generate. That’s a big workload, but how many people are we talking? 50? 500? I’m guessing it’s closer to 50 than 500, but I don’t really know.
So stop doing it then
