3 ms·
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 s
by jfim 7y ago
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.
- dclusin 7y agoDoes 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.
- nindalf 7y agoAny 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?
- jfim 7y agoMost 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.