3 ms·

What would you say is an average engineer tenure in our industry? Roughly speaking, 35% attrition means an average tenure of 2.7 years. When I asked engineers

What would you say is an average engineer tenure in our industry? Roughly speaking, 35% attrition means an average tenure of 2.7 years. When I asked engineers how many year do they think an engineer stays at a job in the industry, I got answers between 2 to 3 years. While I aim for longer relationships and lower attrition rates, knowing the benchmarks is helpful when judging whether something is good or bad.

Frost1x 4y ago
This approach results in a bit of a self-fufilling prophecy, though. If you aim for meeting averages (anecdotal or not), then why would you expect turnover better than average? You'll probably end up optimizing just above that average and likely not hit a critical point high enough to deviate form the underlying problem, because your goal isn't about minimizing turnover even if it's an idealistic goal, when push comes to shove its about getting ROI above all and everything else is secondary. I argue that aside from TC optimizing engineers, most engineers do not want to jump ship that often, especially in the current hiring process. Unfortunately, there's a business culture that created these environments where turnover is this high, sometimes higher. If you provide no stability for your engineers, no WLB, poor internal growth opportunities relative to market, poor training, and grunt work then you should expect this. The way to reduce employee turnover is to provide a significantly better overall deal to employees, otherwise you'll continue to have flighty engineers and just point at the numbers, "well this is what's going on in the industry." Yes, and if you've adopted the rest of the culture of the industry, why would you expect yourself to be much different then? Don't start with the expectation engineers will leave in 2.7 years and build a culture around that, look at places engineers stay longer and build a culture around that. Issue is that those places tend to be less efficient and profitable for investors because the engineers are getting a bigger slice of the pie in some fashion, be it improved WLB, better training, less hostility, etc.
- koliber 4y ago
  I'm not saying that 2.7 years is great. But it is not horrible either, as parent categorized it.
akhmatova 4y ago
A median of about 1.7 years, by my math.
- koliber 4y ago
  I was sharing the average (mean) not median. Can you share how you arrived at that number? I've been trying to verify my approach to this. What I did: - Create a spreadsheet for a simulation - col 1 has "number of people remaining". I start this at 1000 - col 2 has "month number" - col 3 has "% of people who left this month" - col 4 has "number of people who left this month" - in each new row, I subtract the col 4 value from the col 1 value to come up with a new headcount - I fill this down for 10 years (120 months) - for each month, I multiply the number of people that left \ the number of months they were there - I sum the man-months worked, add the man-months for the people who are still there, and divide by 1000 (the number of people in the experiment). This is the average number of man-months.
  - koliber 4y ago
    I just checked the median using my methodology and got 1.95 years. That's closer to your 1.7 than the mean of 2.7 I posted earlier.

Frost1x 4y ago

This approach results in a bit of a self-fufilling prophecy, though. If you aim for meeting averages (anecdotal or not), then why would you expect turnover better than average? You'll probably end up optimizing just above that average and likely not hit a critical point high enough to deviate form the underlying problem, because your goal isn't about minimizing turnover even if it's an idealistic goal, when push comes to shove its about getting ROI above all and everything else is secondary. I argue that aside from TC optimizing engineers, most engineers do not want to jump ship that often, especially in the current hiring process. Unfortunately, there's a business culture that created these environments where turnover is this high, sometimes higher. If you provide no stability for your engineers, no WLB, poor internal growth opportunities relative to market, poor training, and grunt work then you should expect this. The way to reduce employee turnover is to provide a significantly better overall deal to employees, otherwise you'll continue to have flighty engineers and just point at the numbers, "well this is what's going on in the industry." Yes, and if you've adopted the rest of the culture of the industry, why would you expect yourself to be much different then? Don't start with the expectation engineers will leave in 2.7 years and build a culture around that, look at places engineers stay longer and build a culture around that. Issue is that those places tend to be less efficient and profitable for investors because the engineers are getting a bigger slice of the pie in some fashion, be it improved WLB, better training, less hostility, etc.

koliber 4y ago
I'm not saying that 2.7 years is great. But it is not horrible either, as parent categorized it.

akhmatova 4y ago

A median of about 1.7 years, by my math.

koliber 4y ago
I was sharing the average (mean) not median. Can you share how you arrived at that number? I've been trying to verify my approach to this. What I did: - Create a spreadsheet for a simulation - col 1 has "number of people remaining". I start this at 1000 - col 2 has "month number" - col 3 has "% of people who left this month" - col 4 has "number of people who left this month" - in each new row, I subtract the col 4 value from the col 1 value to come up with a new headcount - I fill this down for 10 years (120 months) - for each month, I multiply the number of people that left \ the number of months they were there - I sum the man-months worked, add the man-months for the people who are still there, and divide by 1000 (the number of people in the experiment). This is the average number of man-months.
- koliber 4y ago
  I just checked the median using my methodology and got 1.95 years. That's closer to your 1.7 than the mean of 2.7 I posted earlier.