Comment by noonenohow | Hacker News Reader

Comment by noonenohow | Hacker News Reader

Parent Full thread

> It _is_ important to note that the average retention at Amazon (even for excellent people) is ~50% for 1 year, and ~20% or lower for the 2nd year.

Your comment on "retention" is factually incorrect. I'll presume your comments are in relation to "tech roles" such as SDE SDM TPM etc.

First the number of Amazon hires in "tech roles" increases by ~50% per year. i.e., in year 1 there are 1,000 SDEs hired, year 2 is 1,500, year 3 is 2,250, year 4 is 3,375 and so on.

"Attrition" represents the number of individuals who leave for any reason. This may be because the employee terminates employment or the company terminates employment. Attrition levels are comparable to the rest of the tech industry, ~15% of the current population per year.

"Tenure" is the length of employment. Because hiring greatly exceeds attrition you should expect a relatively low average and median retention. As I recall Amazon is approximately 12 months median tenure. This is comparable to Google at ~13 months, for example.

Lastly the employees current tenure does not have a substantial impact on the probability of their attrition. Indeed the average tenure at exit is somewhere around 3 years. Which, again, is roughly in line industry standards.

In short, the average tenure at Amazon is low because they hire a lot. And while retention is not "good" it is a far cry from what you've represented.

Edit: The gross growth rate is actually above 50%. I've attempted to simplify the values where it does not make a material difference to the refutation.

> Your comment on "retention" is factually incorrect.

I disagree.

> First the number of Amazon hires in "tech roles" increases by ~50% per year. i.e., in year 1 there are 1,000 SDEs hired, year 2 is 1,500, year 3 is 2,250, year 4 is 3,375 and so on.

This could not possibly be true. If it was, they would be hiring ~1.5M technical people a year (since they've been in business since 1997). Perhaps you meant since 2005, but that would still represent 58k people per year hired. According to this article (http://www.geekwire.com/2015/huge-growth-amazon-reaches-2224...), they only have 24k in the state of Washington, where the vast number of people are. So nothing about this statement is true.

> "Attrition" represents the number of individuals who leave for any reason. This may be because the employee terminates employment or the company terminates employment. Attrition levels are comparable to the rest of the tech industry, ~15% of the current population per year.

Can you show me that data? about 15% per year? I've seen vastly different numbers internally, but would prefer not be sued in sharing them.

> "Tenure" is the length of employment. Because hiring greatly exceeds attrition you should expect a relatively low average and median retention. As I recall Amazon is approximately 12 months median tenure. This is comparable to Google at ~13 months, for example.

Correct, but these numbers are heavily obfuscated due to part time hires and vendors.

> Lastly the employees current tenure does not have a substantial impact on the probability of their attrition. Indeed the average tenure at exit is somewhere around 3 years.

Again, show me the data. I've seen the internal stuff. If you'd like not to believe me, feel free; I know I don't believe you.

First "id est" is used to explain or clarify, it is not a list of literal examples.

I'll again state that I am referring to "tech role" workers commonly represented by an "SDE"; not vendors, contractors, hourly, warehouse associates, etc. I am also referring to this population across all of Amazon; Consumer ("Retail"), AWS, and "Digital" aka "Kindle". I make no specific claims to particular internal organizations which may or may not represent the average.

I don't think that anyone is going to repeat or cite a companies internal data. Maybe we can use public data and some simple deductions to arrive at the most plausible explanation.

Your linked "geekwire.com" article is actually an illustration of the SEC 10-K filings. Under that documents definition of "Employees" it specifically terms this to be "full-time and part-time employees." Contractors, vendors, and temp workers are _not_ included in those numbers: "Additionally, we utilize independent contractors and temporary personnel to supplement our workforce." Unfortunately, that does coalesce "tech roles" with distribution ("warehouse") facility workers etc.

Additionally by reading these 10-Ks you'll note a section on "Stock-based compensation." As many have noted RSUs are a significant portion of Amazon compensation. From 2008 annually this is $275M, $341M, $424M, $557M, $833M, $1,100M, $1,500M, & $2,100M through 2016 Q1. Regardless of employee count the cost of issuing RSUs increased by 25-40% per year, with an increase in rate during recent years.

As you, and others, have noted the Amazon compensation package weights the initial RSU vesting schedule to years 3 & 4 of employment. I believe the typical structure is that only 5% of grants vest after the first 12 months, and an additional 15% at 24 months.

Let us use the public data which shows a distinct increase in employee count and compensation costs starting around 2011. Suppose again the company might have had 5,000 employees in tech roles at that time. Using my posited net growth of ~50% we could expect ~25,000 employees after 4 years, i.e. in 2015. Which happens to align pretty well to your citation of 24,000 in WA as of mid 2015. Don't forget that Amazon is a global company, so while a majority of tech roles may be WA based, it is certainly not all. Feel free to decrease the rate to 40% net, it merely changes the target dates by ~1 year.

We can also compare the public data on employee count and stock based compensation. Using the hypothetical ~25,000 tech role employees for ~2015 and $2,100M of stock we arrive at an average of ~$84,000 vested RSUs per tech role employee. Of course RSUs arent a uniform distribution, but that seems to map to self reported salary data. Adjusting for the employee count at 2015 Q2 and the stock based comp cost at 2016 Q1 further improves the fit.

From your other comment you appear to posit that employee attrition is approximately exponential decay each year. I believe you have also stated that the average tenure at exit time is 1 year. In that case employees should have only vested ~5% of their total RSU grants at time of exit, forfeiting 95% on average. How do we reconcile that with the public data showing a ~40% annual increase in the cost of stock based compensation?

Further your exponential attrition should lead to asymptotic employee counts unless offset. I do not believe anyone is claiming a net decrease in Amazon tech role employees, most cite a rapid increase in active employees. What hiring rate do you posit to support both your claimed 50% attrition rates and the significant net increased employee counts?

PS: AMZN 10-K filings for your perusal http://phx.corporate-ir.net/phoenix.zhtml?c=97664&p=irol-sec...

You did not say roles were increasing at 50% per year as an example - that was your stated fact.

I'm saying unless they had 10 employees in 2000 (wrong) or only started the 50% hiring last year (so we have no longitudinal data to go on), there is no number where that level of exponential growth of employees wouldn't result in either a) a company that was bigger than Walmart in employees in just a few years or b) MASSIVE firings. Maybe that's what you're saying? They hire that many people and remove them from the company that fast? I know the removal from the company is certainly correct.

As far as the increase in costs of RSUs, there is a perfectly alternative solutions to your proposal: their stock has increased by 10x in 10 years. So in that way, they're EXACTLY keeping track (https://finance.yahoo.com/echarts?s=amzn+Interactive#{"range...), and not increasing people collecting RSUs at all. You'll note that you list amount of money required for the RSUs, not number of shares. (This is, by the way, another way that Amazon misleads their employees - your yearly cash bonus actually goes down when the value of the stock increases).

So by that logic, they have, in fact, not kept pace with the rise of their stock. New employees are receiving substantially less new grants than older ones (bordering on zero, which can't be true), or they're being removed before they vest.

Further, the 24k in Washington also include many thousands of non-tech roles (Amazon Fresh warehouse in Redmond, Customer Support in Bellingham, some portion of the more than 2500 sales people for AWS, etc etc), so your math does not add up for adding new 25k tech roles, even if we multiply by 40% as a reduction. Doing some approximations using LinkedIn (https://www.linkedin.com/vsearch/p?keywords=amazon&f_CC=1586...) they do appear to have ~24k employees in Washington, but only ~8k in engineering (they appear to add another 4k engineers in India).

Again, these are all public numbers, so they are inaccurate, but they should be good enough to substantially disprove your point. So, without question, your statement about increasing hiring by 50% y/y has no facts to back it up.

As far as active hiring, that is absolutely correct. They are extremely active in hiring - but they have to be because people leave so fast that unless they do, they'd be out of people.

Reply on news.ycombinator.com