I've always been a bit sad that there hasn't been more empirical work on Dunbar's Number. It looks like attempts to validate the modern number range from around

I've always been a bit sad that there hasn't been more empirical work on Dunbar's Number. It looks like attempts to validate the modern number range from around 150 to around 250 (Bernard–Killworth, who seem to be leaders here), but of course that variance has a lot to do with exactly what counts as a 'connection'. You're very right about the mistake of looking only at company size; Dunbar suggested that maintaining the full connection set would take a very large amount of socialization time, and most people have at least some connections outside the office. Even anecdotally, there are lots of people who describe feeling disconnected at work because they spend more time than coworkers on non-work connections. This seems to have been ignored by a lot of people using the number. Wikipedia claims the Swedish tax service limits offices to 150 people based on Dunbar, but that could easily be 50 people too large for a workplace community. Malcolm Gladwell, for whatever his data is worth, claims that Gore-Tex hit cultural problems as offices grew beyond 150. If those anecdotes are accurate, it implies offices function well with high rather than total connectivity. Intuitively, that seems quite reasonable. Cross-organization connections are valuable, but not every pair of jobs has equal value-from-connectivity. And small teams are infamously hard to build because everyone involved has to get along and agree on how to function. If you aim for 90% connectivity (or 50%, who knows) instead, it's possible to maintain a high-empathy organization while still allowing for personality conflicts.