3 ms·
While this is possible, it's relatively unlikely. Top performers can generate enormous amounts of value for a company. As a hiring manager, I absolutely believe
by nilkn 6y ago
While this is possible, it's relatively unlikely. Top performers can generate enormous amounts of value for a company. As a hiring manager, I absolutely believe in the existence of 10x engineers. The reason the term has been so mired in controversy is because folks think the 10x part refers to raw coding ability, but it doesn't. It refers to soft skills, communication, leadership, domain knowledge, business acumen, customer relationships if relevant, and dedication -- overall impact on the company. It's actually quite easy for one developer to have 10x as much positive impact on a small company compared to another.
As a hiring manager, though, I'm stuck in the middle between developers and executive leadership. Executives tend to form a notion of what different roles should be paid, and they just won't bend, even if it means losing talent and spending $1 million+ replacing them over time. The problem is that replacement cost is abstract and easy to ignore, whereas paying a single developer $500k/year to retain them feels very real. Never mind the fact that losing them means multiple projects will be derailed, you'll spend a year looking for a replacement to hire at twice their salary, that replacement will take 3 additional years to reach the same level of domain expertise, it turns out that replacement never takes on the same degree of leadership or earns the same level of trust from colleagues, and in the midst of all of this the absence of your top performer causes two other talented engineers to leave in the following year.
$500k starts to look reasonable in retrospect, but you'll never be able to sell it, so you'll do the above dance instead and spend more money in the end.
- WrtCdEvrydy 6y agoThe issue is we're not able to quantify the costs of technical debt and lack of domain knowledge. We rewrote something in React and adding a feature to it took 2-3 hours but doing the same to the legacy Java code took 3 weeks. This cost was easy to quantify and strengthened proposals to remove the legacy Java code.