Hacker News Reader

Comment by michaelyoshika | Hacker News Reader

On one hand, I don't understand why they need so many employees for a ride-sharing platform. On the other hand, I can understand that people are just so eager to find another growth point so anything looking like a growth point could get almost unlimited capitals then hire a lot.

View on HN

This is a common trope on HN that's strange to me.

I feel like people are unable to see the complexity of an organization from the outside, but normally accept the complexity and need for employees while on the inside.

Might have something to do with the paycheck you're receiving regularly while on the inside...

Or just a better understanding of the organization and the problems they face. The inability to use that understanding and apply it elsewhere is what's strange to me.

"It is difficult to get a man to understand something, when his salary depends on his not understanding it."

Uhh that doesn't really apply here.

I'm sure executives could make a lot of money if they could get the same output with 20% of the workforce.

I'm not the one convincing people my services are needed. I'm the one being poached because they are.

This is needlessly personal, bordering on ad hominem. I've never received a paycheck from Uber but I have worked for large organisations. 3800 people on engineering doesn't seem unusual to me. I can't justify the employment of each of those 3800 but I know for sure that this product could not have been built by 100 engineers.

Sometimes an organization is complex because it has a lot of employees, rather than needing a lot of employees because it's inherently complex. You see the same thing with large software projects.

Sure, there's always going to be bloat, but complexity at scale isn't just because there are a lot of employees.

Twitter is a great example. I bet they have 20+ people working on a specific anti-abuse system. Another 20+ working on a specific ad type that makes the company money. Another team that just works on account security. Another team that works on internal tooling. Or the shared UI components team. Etc until you get the current state of Twitter.

All of those teams don't need to exist when Twitter is small, but do when they scale.

We can definitely agree that as a company grows, it needs more employees. I think the point at issue is that there is such thing as having too many employees. There are multiple opposing forces in tension here, and the tricky part is striking the right balance. I think that GP was trying to say that they hired too many people, and introduced complexity.

A layoff is certainly a sign that they hired too much, I agree.

I was going off what I figured the poster was getting at: they don't understand how Uber has even 80% of their employees.

Is that what they were getting at? It seems to me like you made up that 80% number just now, to prove your point.

I think you forgot what thread you're in.

Uber is discussing laying off 20% of employees. 100 - 20 = 80.

It's completely reasonable to assume the OP was trying to figure out how they have so many employees, even after the layoffs.

But it's not about employees. It's about capabilities, and working backwards.

eg look at datastores. If you can fit your data in postgres on a single (potentially really expensive!) computer, you probably need two or three of those computers for a primary plus a couple replicas and a DBA.

Once you outgrow that, going distributed costs a team. There is a big elbow in the cost and headcount curve.

Once you want a different specific capability beyond the standard distributed datastores, that costs multiple teams. Yet another big elbow in the cost and headcount curve.

This happens on data stores, applications, translations, currency, etc etc etc. Imagine a Greek person with a Greek bank account driving Uber in France being paid with an American amex card.

I talked to a friend of a friend working there. It sounds like they previously hired engineers before having projects they needed them for. It's fairly clear some of these companies just set targets for growth and hire people to match, rather than to meet some fundamental business demand.

Or people just see Uber the app and think to themselves why is that so hard. All the while ignoring / not seeing all the other pieces that are needed to support that app. And this is just from the technology perspective.

I don't have any insider knowledge, but I imagine just the support and operations org to manage and respond to all those riders and drivers can't be small.

>This is a common trope on HN that's strange to me.

It's because HN is not filled with people with knowledge of the industry anymore. People in tech are now the minority.

You must be joking.

Given the quality of discussion on this website, no im not.

It's the classic line: "That's so simple, I could have built that" which warrants the classic response: "fuck you, you didn't".

Why the hostility?

Sure we have the benefit of hindsight to solutions that fit within the product demand fitness landscape. But by identifying areas we could have attacked, we shape our intuition and gain the confidence to exploit other areas of opportunity. That's valuable.

I think this warrants a reframing - instead of focusing on missed opportunities, look ahead for new ones. There will be gradients for the foreseeable future. Learn to spot them, learn the skills you need, and then position yourself to solve the problems.

With the knowledge I have now, I could have built Facebook, Twitter, Uber. But I didn't know to do it then. I'm trying to put myself in the right place to solve problems that matter to me that I also think could move the needle.

It’s faux/facetious hostility. And it’s I response to a naive/offensive/belittling comment.

I’ve not worked in a company as big as Uber, but just in mid sized companies with also midsized markets and there is never an end to the work that needs done. I don’t see why it’s hard to believe it takes a lot of hands to build something like Uber.

And I think you’re wrong that you could have built Uber/Twitter/facebook even with what you k ow after the fact. You could have built a toy simulacrum, but put it before millions of users and it’d fall apart, and not just from scale, but missing features; all the different OSs, regional controls, payment controls, mapping, pricing models, etc.

If there was a spec for one of these companies it’d be thousands of thousands of pages long and most of it you don’t know about; no single person could.

So a little faux hostility seems warranted.

People tend to subconsciously assume the operations of an organization are linearly associated with the public-facing products and services they provide. People quantify based off of what they know and understand. Nobody can quantify the effort associated with an effort they didn't know existed.

I seen plenty of people who think their orgs are bloated. I know I've felt some teams could lose half the workforce at no real cost.

That's the case for many tech companies, why does Twitter need thousands of employees?

Turns out there's lots of employees to do all kinds of things, the app side is only the tip of the iceberg. There needs to be people collecting metrics and analyzing them, people building the platform that employees use to deploy code and run batch jobs, maintaining infrastructure (IIRC Uber did onprem -> cloud -> onprem), people interacting with customers and drivers, local expansion teams, and so on.

Does twitter really need those metrics? Just because people want metrics doesn’t mean they need to or should be collected. I have seen over the years posted here and again that highly targeted advertising and the ensuing surveillance infrastructure that that entails hasn’t really resulted in a significant improvement over non-targeted generic advertising like keywords and brand based advertising. So they could probably be as effective from revenue perspective without all the infra to spy on people and the manpower that requires.

Huge revenue slumps like these tend to facilitate come to Jesus moments with regards to staffing and mission priorities. I suspect if twitter had or was going to losE confidence of the stock market and needed to show improvement a lot of the so called mission critical metrics would probably go.

Any change deployed by large software firms are checked to make sure that no team wide and company wide metrics are adversely affected. How do you do that if you don't have the metics in the first place?

Most metrics aren't really related to advertising or personal data collection, though. Things like monthly active users (part of the quarterly shareholder report), conversion rates, cart abandonment rates, feature usage, active minutes, and so on are all part of the usual development process.

These are used pretty heavily when launching new features or changes to existing features, in order to make sure that no key metrics are adversely impacted. For example, if one launches a new feature but users are confused by it, they might be less likely to succeed at completing a transaction, thus negatively impacting revenue. The change will then be re-evaluated or even rolled back.

Think about landing at an airport in the before times and getting an Uber. You don't just put a marker on a map view, you have to be directed to where the airport set up ride share pickups. The drivers who want to wait in a queue to pickup airport riders need to be told where they're allowed to wait. There are millions of little edge cases like this, all with different, non-standard laws in every city they operate in. It's not just running a big server that keeps a table of pickup coordinates.

According to Price's Law 50% of the work is done by the square root of the number of people. As the amount of stuff you need to do grows linearly, the amount of people you need to do that stuff grows exponentially.

I think you mean quadratically.

In popular usage of the term, anything above linear is “exponential” unfortunately.

Price's Law is great. I think it provides insight into so much of how the world works.

I am curious to see how Red Hat performs now that it's been swallowed up by IBM.

Is there any empirical evidence of Price’s law? Would love to see it.

I could only find this study (admittedly from 1986) that disproved the original hypothesis - https://www.sciencedirect.com/science/article/abs/pii/030645...

Their ride sharing platform is very sophisticated, though. Pathfinding based on gps locations, maps, traffic, and other obstacles, and solving traveling salesman problems with constantly changing variables are generally not simple problems (every ride sharing platform does it, sure, but that doesn't mean it's easy, or that they all do it well). Plus they are at a very high scale and high reliability, both of which require large amounts of employees (things don't just scale themselves). On top of that they have a lot of vertical / horizontal integrations.

They are powered by Google,so travel routes, traffic info and all other difficult things come from them.Not saying it's easy, but main challenges are definitely somewhere else.

None of the above are powered by Google. And marketplace difficulty comes from regional constraints, product differentiation, and incentives to try and achieve equilibrium (e.g. demand forecasting). The resources required for these are massive.

From people I’ve spoken with, the only mapping thing google provides that they use is the base map and road segments. In some markets they use Open Street Maps or TomTom when the data is better than Google. All the routing tech is their own and works atop these three data sources.

https://www.google.com/amp/s/techcrunch.com/2015/11/12/uber-...

Uber has tons of regional offices for local expansion, lobbying, marketing, loans, etc. Not sure what % that adds up to but I can imagine they have many many employees in that category internationally.

Read up on Uber's infrastructure, and you will be quick to reconsider.

TL;DR, every marginal employee generates enough marginal revenue at scale that they more than pay for themselves, and so it's worth it for the company to keep adding people (at least while the business is doing well)

How else would they re-write all their "data science" code from python to go?

Reply on news.ycombinator.com