Hacker News Reader

Comment by scottlegrand2 | Hacker News Reader

As a former Amazon manager, I don't have a problem with forced attrition so much as the proportion of my time that was put into finding the candidates for that forced attrition.

There are a tremendous number of low achievers in every industry and if you keep them around they are toxic, every bit if not more toxic than Amazon's culture in 2019.

That said, I would have liked to have seen most of the time dedicated to this task allocated to developing the careers of my direct reports because I think that is of much higher value to any organization.

Not hiring idiots in the first place is IMO the proper solution to this problem, but that now seems intractable. It seems there's a real temptation to scrape the bottom of the barrel to fill positions these days.

View on HN

This makes no sense. If you have low achievers, your job as a manager is to deal with them. It is not as though you can’t PIP someone as needed. Forcing it at a certain rate is just unnecessary and creates terrible incentives.

Presumably the aim of forced attrition is to take managers who are firing grade D performers and retaining grade C performers, and turn them into managers who fire grade C performers.

Of course, it's pretty questionable whether it achieves that aim - one would think it encourages managers to keep incompetent employees around as buffer.

Forcing a certain rate definitely makes sense if you're large enough and have the data to back it. If you're a small dev team with 10 people, it's possible everyone's doing great. If you have 20,000 devs, it's unlikely. You know the % of bad hires over a long enough time period. Forcing the rate often forces managers to disregard emotional connections to make the best business decisions. It sucks but you have to remember a company has no loyalty to you anymore. You should show it the same.

Except in all cases it's a small team. The company with 20,000 devs still imposes this on each individual team. You just have every small team doing it. It's a disaster and it makes no sense.

I very firmly believe this, but I have created zero successful companies, and Jeff Bezos has created the most successful company that the world has ever seen, so there's that.

It does seem like Microsoft became far more successful after they did away with their famously bad similar policy, so maybe Amazon would be even more successful than they already are if they valued their employees a bit ;)

And if his management doesn't support that sort of approach or developing his team or hiring well (support with resources not just verbal agreement)?

> Not hiring idiots in the first place is IMO the proper solution to this problem,

If you hire all rockstars and you have to let go of 10%, then how does this help?

MS managers had a hack around this - a few months before expected lay-off cycle they hired some incompetent developers or developers from their competition they hated, then let them go at the time of lay-off. Wouldn't be surprised if Amazon managers used similar tricks.

Across thousands of hires, can you hire all rockstars? When you hire 1,000 people, how many do you expect are underperformers?

No matter how many people you hire, at least half of them are going to be at or below the median for your new employees.

So I'd say 500. You'd have to be very good at hiring to get a lower number, and judging from my anecdata, companies that hire software pros use hiring processes that are terrible predictors of future performance, if not entirely useless.

If a company isn't positioned to turn a profit on the output of a rockstar, why is that company trying to retain a development team at all? Every company I've worked at for a decade was eagerly hiring every rockstar they could find, except the one on the verge of bankruptcy that was trying to get acquired.

It's a good tactic, denying the competition the resources, until it gets to expensive.

> If you hire all rockstars and you have to let go of 10%, then how does this help?

It keeps salaries lower than where they would be and pushes the remaining 90% to work very hard.

Really?

Do you not think such a culture is actively detrimental? How can you ask anyone to give 110% if they might be getting sacked next year?

If you have a tremendous number of low achievers, maybe its because the only people you can retain are people who can't find jobs elsewhere fast enough. I know I wouldn't be waiting around for my turn.

> Do you not think such a culture is actively detrimental? How can you ask anyone to give 110% if they might be getting sacked next year?

As much as I dislike the forced attrition mechanism, people putting in 110% are not going to be forced out. It's the people putting in 30% that are going to be the attrition targets.

People putting in "110%" are also more likely to get either denied promotions (managers don't want to lose their most productive assets) or are driven out as major threats to their superiors.

110% and 30% by what metrics though?

My 30% could be somebody else's 110%, though that's not the likely scenario we're talking about.

What if instead, some people put in 110% by putting in 14 hour days and others are judged more harshly who have family or other obligations. This is how we get an industry of 20-somethings with no lives other than work. In a forced attrition scenario, if I were on a team with people significantly younger than myself, I would be shitting bricks.

> if I were on a team with people significantly younger than myself, I would be shitting bricks.

I was noticeably older than most of the people in my team in Amazon, and I was never shitting bricks[1]. The ones not reaching performance targets were always crystal clear to _everyone_. If you can't spot the underachievers, maybe that's the time to worry.

Despite how detailed the Amazon interview process is, people slip through the gaps and you end up with someone who somehow can barely write functional code, let alone maintainable code. Or you find the person is capable of writing code, just maybe not today, or tomorrow. Is the day after good with you? There are much more interesting things to be doing on-line instead of writing code.

[1] I was reminiscing with a co-worker in the kitchen one lunchtime about the respective things we'd had to do to deal with the Y2K bug at the places we'd been working at. Another talented and very capable developer co-worker overheard us and commented "I'd never even touched a computer before, back then."

If a company is hiring people who can't write code then their hiring process has huge problems. In my 15 years of developer experience at 5+ companies, I've never once met a developer who couldn't code. I don't know where companies are finding all these developers who supposedly can't code.

I often think that "programmers that can't code" is an euphemism for "programmer why won't do unpaid overtime to catch up on the code base" or "programmer who has no experience".

The industry is not expecting mechanical engineers to actually do advanced work fresh of university but somehow in some places programmers are thought of as a ready worker package for the sewing machines.

I'm also yet to meet these programmers who can't code. There are plenty of engineers that barely can code though, but they don't call themselves programmers. And there are plenty of inexperienced programmers that can't code and if they won't learn they will drop out for other positions.

>If a company is hiring people who can't write code then their hiring process has huge problems.

There's a difference between "can write code" and "will write code" which is what I was more intending to express in my post.

Everyone that gets a Dev job in Amazon has to be able to pass multiple loops where development skills will be tested.

I've definitely had those dunning-kruger effect moments in my career where I thought I was an underperformer until I got feedback from management to confirm otherwise. In fact, I would say that most companies do not have much of a system at all for positive and negative feedback. It's usually one or the other and done badly.

They will be, eventually. If you constantly fire the ones at the bottom, the only way to stay on top is to put more and more. Which means that at some point, your 110% is going to be the lowest ranked.

No it probably won't be the people giving 110%. But you want those people invested it what they're doing, or they won't be giving 110%.

To put it another way. You should be focussed on maintaining and motivating the top 10%, not on getting rid if your bottom 10%.

I disagree with this. Every organization, and every department has a bell curve of talent. Average and low-average people hire sub-par and poor talent because everyone thinks they're rockstars, but statistically that cannot be true.

"everyone thinks they're rockstar"

Exactly, but knowing that, even if you think you're safe, you might not be, so what do you do?

Plus you're saying below average people hire below poor talent.

This just means that really the good manager is sacking good people, whereas the poor manager is sacking just 1 of his limitless supply of poor hires.

So as an average employee what do you want, good manager and be bottom of the bell curve, or poor manager and be top of the bell curve?

Huge oversimplification. Some companies can hire very good engineers most of the time. Most companies cannot.

Were you explicitly told to find people on your team(s) to get rid of? Or was in at an org-level? Were there actual quotas?

I have worked at big companies where this happened - one of my co-workers commented that their group found this easy one year as they had two people with terminal cancer.

This isn't as bad as it sounds as they got medical retirement.

Not OP but worked in one of top 10 companies for a senior role and was also involved in hiring tech team.

There was no explicit "We need to get rid of x people" but there were performance reviews every year and getting the lowest performance mean being reassigned to another team and 3 consecutive lowest performance means forced attrition.

I don't even know why it's controversial (unless its a botched up execution). Most companies have dead weights that got in somehow and are toxic to the culture and a burden to the team.

None of my ex teammates in Amazon was literally fired (AFAIK). Many had disagreements with management over the inflated expectations and left within 3-4 years... despite being very good engineers.

We are not talking about 5% or 10% attrition rate here.

What I found to be the biggest problem at Amazon beyond the issue here was not so much inflated expectations but rather top-down technologically backwards micromanagement of how to get things done.

The most flagrant example of this in my experience was when I was ready to hand an AWS customer a simple script that would replace their command line application and magically make it run in the cloud, uploading inputs and, and downloading results, it was blocked from release to the customer immediately because it wasn't a full web app with staffing and an on-call pager.

It's controversial because the bar for 'dead weight' escalates over time. I worked for a company where eventually a Satisfactory rating became inadequate. Just doing your job wasn't good enough.

Lol, I wonder how that was passed down. "I am sorry, but we're not satisfied with your satisfactory performance".

It's controversial, because "lowest performance on the team" doesn't mean "poor performance".

True. And the rating was by a higher manager across multiple teams. And he was under no obligation to give lowest grade to anyone. Most of them got "average" and few of them exceptional or very good etc.

Then you're not describing the same system that Amazon does. The point of that system is that there's a certain fixed percentage of people who will get the lowest grade, no matter what.

I'm curious, though: is it not possible to have a small team of 10 excellent people? And even if all 10 are excellent, one must get the lowest score. Is it really good to always reassign or fire someone based on the lowest score?

It was not a "team" or "project". This company operated on Toyota's model of dual manager (or it's variant). One for the project and one overall. The overall manager oversaw a a fixed number of people (around 200-250) and he could rank few of them the lowest. He is not obliged to and many managers never gave the lowest score to anyone.

Also this manager would get the feedback from all the teams his reports worked for and make his own decision. Quite often he would even ask the employee why his project manager rated him low.

Overall from what I had seen the process worked not too shabby even though it was unfair.

So it would take 3 years worth of poor performances to get let go?

Yes and three consecutive years of lowest performance. This was not set in stone or written in any emails but an HR's email about "Consistent low rating" mean the employee would be on his way out.

The decision and final axing was always done by the HR and managers could only intervene if they thought it was unfair.

The counterargument here is empirical: there are companies that manage to do fine at Amazon's scale without forced attrition. Google is one of the most notable examples.

Ah yes Google, the place where as possibly the world's first or second (full-time) CUDA programmer, with 14 filed patents, and the world's fastest implementations of molecular Dynamics (CUDA ports of Folding@Home and AMBER), I was thrown into blind allocation by Amit Singhal and I was told point-blank by Jeff Dean that GPUs would never serve a role at Google because the Intel processor roadmap was just about to steamroll them out of existence. Yep that's exactly how it played out, right?

Anyway, they have PIPs too. My slapstick comedy of 4 months there was nothing compared to the horror shows others endured there. That's how I ended up at Amazon in the first place and my first four years there were actually pretty good. It wasn't until the New York times article that things started going downhill there in my opinion.

Your response to "companies can thrive without forced attrition" is "Jeff Dean was wrong about a thing once". Does that make sense to you?

I wasn't responding to those exact points but rather to the hagiography and idealization of Google as a place without forced attrition.

Google has great food and great perks, but it was one of my worst employment experiences of my entire professional career. Amazon, warts-and-all, was far better to me and for me.

And I haven't even mentioned the time in 2014 when the manager of Android Store recommendations asked to speak with me and so I showed up on the googleplex, only to have him walk me out of the building to the swamp behind the googleplex to tell me that yes I'm on some sort of blacklist and he would make an exception to bring me in if I were willing to work on these recommendations. Now this is really interesting and nearly industrial espionage because I was secretly working on Amazon product recommendations at the time and he should not have known I had any such experience because we had just won our first web lab. It all ended when I asked him if I could use GPUs for this and he said no.

So my point is just because they don't have forced attrition doesn't make them a great employer across the board because WTF was that about?

I don't understand what your point is. Yes, they have PIPs. No, that's not at all the same thing as forced attrition.

I don't know about this personal drama with two particular programmers. I'm not sure it's relevant.

> It wasn't until the New York times article that things started going downhill there in my opinion

+1 to this. The reaction to that article was, in my personal view, the wrong one.

Do you mean Amazon management reacted incorrectly to it? How so?

As an Amazonian, I think some of the criticism was entirely valid- just not most of it. But the reaction, the changes that occurred, made things worse (again, imho), not better.

What got worse? In terms of reaction to the article by Amazon I'd think positive actions or inaction were more likely than negative actions.

Why did the NYT article have a negative impact?

I heard horror stories about Amazon many years before the NYT article. The idea that Amazon suddenly went downhill afterwards is ludicrous.

but google has an insane hiring bar meaning that they rarely have false positives. amazons philosophy is hire fast fire fast. it’s too late for them to raise their bar considering a lot of their employees wouldn’t reach it

> amazons philosophy is hire fast fire fast

Oh god, I can't disagree more. I've been an interviewer at Amazon. I've probably done 50+ interviews in house and hired maybe 5 of those. I'd really prefer if we'd lower the hiring bar and focused more on building (and RETAINING) engineers. It would be far less effort, imho.

that’s because 90 pct of people applying to any job are unqualified. all three times i’ve interviewed at amazon getting an offer was trivial.

Sounds right. A friend who worked there said he did 100 interviews for devs, with 0 ending up hired.

Not at all. I've been in many interview loops in Amazon. The hiring bar obviously depends on the team but it's usually higher than other FAANG without Google

Don't want to ask anything inappropriate, but how much of your time did it end up being? I remember they had some some kind of value where they want "every hire to be better than the last one" or some platitude.

Seems like forced attrition is .. not ideal. Having some consistent way of assessing performance of engineers across teams, and then looking at the front end (hiring groups involved in creating the low performing teams) and on management (focus on who is accumulating higher numbers of low achievers) would be better.

But having seen excessively stable R&D organizations, I completely agree that not getting rid of low achievers is extremely toxic to a high performing team (product quality and morale).

>Not hiring idiots in the first place is IMO the proper solution to this problem,

Presumably leetcode avoids this. what is your opinion about that? Is it effective in your experience?

Leetcode is something you can study for a couple of months and use fly through the technical portion of an interview. It is also something that good engineers with a decade of experience can totally fail at if unprepared to the same degree.

Leetcode questions measure interview study and how far removed you are from college classes, not how competent you are at your job.

Being good at leetcode means you are good at spending plenty of day hours on interview prep and not real work.

Would you really want to hire such a person who does no work all day, and only prepares to move on from current company?

...yea except every dev in the game knows thats where to study.

Indeed, except in 2010 it was topcoder, and they even told me upfront to study it, so I did, for the two months preceding the interview.

Several of my interview questions were slight variants on recent topcoder questions. And other than annoying the guy who clearly had some complicated algorithm he wanted me to write that I instead simplified to the quadratic equation, things went well.

Reply on news.ycombinator.com