4 ms·

Without meaning any offense to anyone, I am also genuinely curious how is a company such as Twitter have so many employees? Outsiders like myself know what goes

Without meaning any offense to anyone, I am also genuinely curious how is a company such as Twitter have so many employees? Outsiders like myself know what goes into running GE or Tesla (and is easy to understand), but I would really appreciate someone from silicon valley enlighten about various needs for a company such as Twitter (or WeWork, etc.); what really goes on inside the doors, how are teams organized, what are some of the most challenging problems? If I were running Twitter, I would freeze the feature creep, have a robust security and site-reliability team, perhaps make small improvements to UI/UX by eliminating friction/impedance. Similar to how Craigslist and surprisingly how Ebay is today - it is pretty much the same experience as 2008, I vividly remember. Instead, engineering resources can be channeled to new ventures (or augmentation projects) for additional growth.

momokoko 7y ago
Because at scale small changes can save or make a company more money than an engineers salary. Simply making a process 2% more efficient can save hundreds of thousands of dollars a year. Or increasing advertising conversions by 1%. etc etc.
- bob1029 7y ago
  Right, but system creation+enhancements can be made by a single engineer in a large % of those cases. Overwhelmingly so if you already have a working live system to start with and you are seeking incremental improvements (vs a full rewrite). There aren't many technical challenges I have encountered, when reduced to their smallest functional dimensions, that couldn't be fundamentally addressed by a single engineer. Obviously, efforts like writing an entire (practical) operating system are impossible for a single engineer to accomplish, but even then a single engineer could lead the entire effort or massive parts of the effort (e.g. the Linux kernel). Conversely, I have seen far too many cases of "too many chefs in the kitchen" when it comes to a software project. This is the end of the spectrum where you end up stuck on Docker with a stochastic and otherwise insane architecture that isn't even consistent within each product, much less across the entire organization.
  - farisjarrah 7y ago
    Most of the time one competent 10x engineer could indeed build out all the special features that a company needs. The problem is that one person generally doesn't know all the best practices required of every problem domain, so they may implement things poorly. Furthermore, the one dev thing really doesn't scale up. As soon as your product gains any critical mass and you have an outage then the platform is hosed because no one knows how to fix anything except MAYBE that one engineer who implemented all the special features.
    - B_Throwaway 7y ago
      There are no systems running Windows 10X in production yet.
      - legostormtroopr 7y ago
        Its a shame this amazing reference is being downvoted. Hopefully Windows 10X puts an end to the "10x developer" meme.
      - dtparr 7y ago
        Can't tell if you're trolling, so in case you've confused others. http://svdictionary.com/words/10x-engineer http://svdictionary.com/words/10x-engineer
  - nostrademons 7y ago
    The percentage of changes that actually improve things drops significantly as the company gets more optimized, so you need to try more things to have a good chance of actually getting one implemented. Say that a tweak to the page improves ad revenue by $50M. 98% of tweaks have neutral or negative effects. The expected value of the tweak is thus $1M, well over one engineer's salary. You need to have 50 of them going on in parallel to have a good chance of actually finding a successful tweak, though. (Well, technically about 150 to have a 95% confidence that at least one will prove useful, but there's also a good chance that more than one will be useful then.) That's what all the engineers are doing - independently implementing different changes and discarding 98% of the results. When I did this kind of work, the vast majority of engineering time was actually spent measuring the results. It was not unusual for the actual implementation work to be done in about 3 hours, but the experiment design, approval, configuration, logging, data-collection, and analysis to take 3 weeks. And that was with a fair amount of existing tooling for measurement & process.
- tanilama 7y ago
  But WeWork doesn't make any money. That is the problem. No matter how many smart engineers they hire, it is just adding additional cost, until they start to turn a profit. Majority of its business value goes to real estate owners.
stygiansonic 7y ago
Even though this article's title sounds like it, this isn't meant to be flippant: https://danluu.com/sounds-easy/ https://danluu.com/sounds-easy/ Also see the other child comment: Once you have a significant amount of revenue, investing the effort improve things by 0.1% (or even 0.01%) can have a not-insignificant impact on the bottom line.
darawk 7y ago
I have always wondered this too. On the surface, Twitter doesn't seem like a very elaborate service. I'm sure there's serious thorny scaling problems to deal with. But how many engineers does that really take?
- olivermarks 7y ago
  Remember the fail whale? Scaling is hard.
  - kevin_thibedeau 7y ago
    It's hard when you build a global business on a scripting language.
acchow 7y ago
Twitter has 321 million monthly active users. WeWork has 466 thousand members. The scale is quite different.
Johnny555 7y ago
As a company grows, development becomes harder due to the need to interface with so many teams -- a feature that you used to be able to implement a couple days after a lunchtime discussion with your coworkers turns into a detailed design and design review, design revisions to prevent breaking functionality that other teams use, a revised design review and approval, security review and approval, and then finally implementation, but you find that in the 3 months since you started the design, the functionality you were planning on using was changed by yet another team and the affect on your feature was missed during that project's design review. Plus, the problems you need to solve get more complicated, you have new compliance needs, paying customers are demanding new features that may require changes across your entire stack, performance is more of an issue (saving 3% of your compute costs didn't matter when you were paying $500/month for infrastructure, but saving 3% of 10 million dollars/month starts turning into real money) and writing and deploying software used by hundreds of millions of people is literally orders of magnitude harder than software a few million people use.
- spectramax 7y ago
  I found this from another front page article about Automattic: https://automattic.com/about/ https://automattic.com/about/ If you go to their about page, they show a little table, where they have 163 million monthly active users and only 966 employees. Not sure how good this comparison is with Google and Amazon because the kind of tools Amazon builds vs what Automattic builds (WooCommerce, Wordpress) are totally different.
bravoetch 7y ago
So you'd freeze the feature creep, but keep working on the set of features you listed? I think you answered your own question.
danso 7y ago
One of Twitter's main challenges at this stage is content moderation and fraud detection, a technical challenge for which solutions are not straightforward or easily scalable.
Spooky23 7y ago
Consumer VC companies are businesses that require growth at all costs. So they need to invest in man-hours to churn out features. If they were a normal company, they’d focus on the top x projects that generate value and spend as little as possible on the rest.
CydeWeys 7y ago
Twitter runs an entire advertising platform. If you've ever logged into Adwords or Adsense you'll know how complicated that is. I wouldn't be surprised to learn that the basic twitter.com tweeting/commenting/retweeting system isn't remotely the most complex system that Twitter even has.
tootie 7y ago
8 people work on the tweeting part, 4000 on ad sales.

momokoko 7y ago

Because at scale small changes can save or make a company more money than an engineers salary. Simply making a process 2% more efficient can save hundreds of thousands of dollars a year. Or increasing advertising conversions by 1%. etc etc.

bob1029 7y ago
Right, but system creation+enhancements can be made by a single engineer in a large % of those cases. Overwhelmingly so if you already have a working live system to start with and you are seeking incremental improvements (vs a full rewrite). There aren't many technical challenges I have encountered, when reduced to their smallest functional dimensions, that couldn't be fundamentally addressed by a single engineer. Obviously, efforts like writing an entire (practical) operating system are impossible for a single engineer to accomplish, but even then a single engineer could lead the entire effort or massive parts of the effort (e.g. the Linux kernel). Conversely, I have seen far too many cases of "too many chefs in the kitchen" when it comes to a software project. This is the end of the spectrum where you end up stuck on Docker with a stochastic and otherwise insane architecture that isn't even consistent within each product, much less across the entire organization.
- farisjarrah 7y ago
  Most of the time one competent 10x engineer could indeed build out all the special features that a company needs. The problem is that one person generally doesn't know all the best practices required of every problem domain, so they may implement things poorly. Furthermore, the one dev thing really doesn't scale up. As soon as your product gains any critical mass and you have an outage then the platform is hosed because no one knows how to fix anything except MAYBE that one engineer who implemented all the special features.
  - B_Throwaway 7y ago
    There are no systems running Windows 10X in production yet.
    - legostormtroopr 7y ago
      Its a shame this amazing reference is being downvoted. Hopefully Windows 10X puts an end to the "10x developer" meme.
    - dtparr 7y ago
      Can't tell if you're trolling, so in case you've confused others. http://svdictionary.com/words/10x-engineer http://svdictionary.com/words/10x-engineer
- nostrademons 7y ago
  The percentage of changes that actually improve things drops significantly as the company gets more optimized, so you need to try more things to have a good chance of actually getting one implemented. Say that a tweak to the page improves ad revenue by $50M. 98% of tweaks have neutral or negative effects. The expected value of the tweak is thus $1M, well over one engineer's salary. You need to have 50 of them going on in parallel to have a good chance of actually finding a successful tweak, though. (Well, technically about 150 to have a 95% confidence that at least one will prove useful, but there's also a good chance that more than one will be useful then.) That's what all the engineers are doing - independently implementing different changes and discarding 98% of the results. When I did this kind of work, the vast majority of engineering time was actually spent measuring the results. It was not unusual for the actual implementation work to be done in about 3 hours, but the experiment design, approval, configuration, logging, data-collection, and analysis to take 3 weeks. And that was with a fair amount of existing tooling for measurement & process.
tanilama 7y ago
But WeWork doesn't make any money. That is the problem. No matter how many smart engineers they hire, it is just adding additional cost, until they start to turn a profit. Majority of its business value goes to real estate owners.

stygiansonic 7y ago

Even though this article's title sounds like it, this isn't meant to be flippant: https://danluu.com/sounds-easy/ https://danluu.com/sounds-easy/ Also see the other child comment: Once you have a significant amount of revenue, investing the effort improve things by 0.1% (or even 0.01%) can have a not-insignificant impact on the bottom line.

darawk 7y ago

I have always wondered this too. On the surface, Twitter doesn't seem like a very elaborate service. I'm sure there's serious thorny scaling problems to deal with. But how many engineers does that really take?

olivermarks 7y ago
Remember the fail whale? Scaling is hard.
- kevin_thibedeau 7y ago
  It's hard when you build a global business on a scripting language.

acchow 7y ago

Twitter has 321 million monthly active users. WeWork has 466 thousand members. The scale is quite different.

Johnny555 7y ago

As a company grows, development becomes harder due to the need to interface with so many teams -- a feature that you used to be able to implement a couple days after a lunchtime discussion with your coworkers turns into a detailed design and design review, design revisions to prevent breaking functionality that other teams use, a revised design review and approval, security review and approval, and then finally implementation, but you find that in the 3 months since you started the design, the functionality you were planning on using was changed by yet another team and the affect on your feature was missed during that project's design review. Plus, the problems you need to solve get more complicated, you have new compliance needs, paying customers are demanding new features that may require changes across your entire stack, performance is more of an issue (saving 3% of your compute costs didn't matter when you were paying $500/month for infrastructure, but saving 3% of 10 million dollars/month starts turning into real money) and writing and deploying software used by hundreds of millions of people is literally orders of magnitude harder than software a few million people use.

spectramax 7y ago
I found this from another front page article about Automattic: https://automattic.com/about/ https://automattic.com/about/ If you go to their about page, they show a little table, where they have 163 million monthly active users and only 966 employees. Not sure how good this comparison is with Google and Amazon because the kind of tools Amazon builds vs what Automattic builds (WooCommerce, Wordpress) are totally different.

bravoetch 7y ago

So you'd freeze the feature creep, but keep working on the set of features you listed? I think you answered your own question.

danso 7y ago

One of Twitter's main challenges at this stage is content moderation and fraud detection, a technical challenge for which solutions are not straightforward or easily scalable.

Spooky23 7y ago

Consumer VC companies are businesses that require growth at all costs. So they need to invest in man-hours to churn out features. If they were a normal company, they’d focus on the top x projects that generate value and spend as little as possible on the rest.

CydeWeys 7y ago

Twitter runs an entire advertising platform. If you've ever logged into Adwords or Adsense you'll know how complicated that is. I wouldn't be surprised to learn that the basic twitter.com tweeting/commenting/retweeting system isn't remotely the most complex system that Twitter even has.

tootie 7y ago

8 people work on the tweeting part, 4000 on ad sales.