3 ms·
This probably would mainly work well with A) highly talented software engineers who want to make much less than they could make for easier lifestyle or B) low p
by metacritic12 5y ago
This probably would mainly work well with A) highly talented software engineers who want to make much less than they could make for easier lifestyle or B) low productivity companies.
The reason is that companies that employ software engineers are engines for transforming labor to value. They get a multiple in dollars in revenue out of their labor cost to you. Suppose this multiple is 3x. Then if an software engineer works 80% time, the indifference point for the company isn't paying 80%, but actually paying 40%. I don't think that number is acceptable to most employees.
Of course, this is politically popular with the HN crowd because many people are software employees. But you need to look at it from the other side too.
How could you make it work then? Well if you were a superstar before that made $200, you probably can get away with 4 days as week and $80. Or if the company is a low productivity company (e.g. their multiple of EV on the employees time is really low like 1.1), then they can probably 80% time for 75% pay.
With matching markets (employee/employer) you need to solve both sides' problems. (It's easy to solve just one side's problem: just advocate for doubling/halving employee pay. Employees/employers will love that.)
- lukevp 5y agoI don’t understand your argument. If 8 engineers work for $100 each, work 40 hours per week and output 320 hrs of work a week, or 10 engineers work for $80 each, 32 hours per week and output 320 hrs of work… there is overhead of managing more people, yes, but the people will likely be more productive per hour than the 40 hr per week people. Why does it matter if the company is making 3x or 30x on the hours per employee when we’re discussing the hours/pay ratio per employee, which would stay the same?
- metacritic12 5y agoThat's a good point. It depends on whether the company's production function really is just 3x the cost of each employee. Clearly, on average, it has to be a high multiple, but on the margin it probably does drop to 1x. I concede that the argument for sublinear would have to be something about overhead or n-choose-2 communication costs.