Google Calculating Which Employees Are About to Quit

http://blogoscoped.com/archive/2009-05-19-n33.html

102 83 17 years ago

makecheck• 17 years ago

...the algorithm helps Google “get inside people’s heads even before they know they might leave”...

Sounds like the department of pre-crime. :)

It's fascinating: enough Google employees want to leave, that Google is studying their behavior. Clearly, even amazing perks, challenging work and brilliant colleagues will wear thin on a person? Or maybe it's not that there's anything wrong with Google per se, but that people are regularly trying to branch out and start their own companies, etc. and no amount of pampering will change those ambitions.

amichail2• 17 years ago

Amazing perks? Such perks should be insulting. Who would make a career move based on free food?

As for brilliant colleagues: that could be good or bad depending on whether you get along with them. Sometimes it's better to be the brightest person in the room.

brk3• 17 years ago

* Who would make a career move based on free food*

Well, all the people I've ever worked with need to eat.

If food is provided free (and it's edible), this seems like a perk with a direct measurable benefit.

abossy2• 17 years ago

Perks span a broad range of offerings. "Free candy," for example, isn't all that enticing, but having speakers from all over come speak to establish an environment with strong intellectual values is something that would definitely sway me in a career move.

neilk2• 17 years ago

Switching jobs just for free food would indeed be a bad idea. Gourmet, incredibly delicious food every day is pretty significant. It really brightens your day. Google's cafeterias, if they were not just for Googlers, would easily be among the best restaurants on the peninsula.

The reality: it's way too easy to overindulge, and over time, it's less that they're giving you food and more like they're taking away your independence.

As for your other point, it's good to have a variety in your career -- sometimes, be the smartest guy in the room, other times, the dumbest.

At Google, sometimes you find youself in a group of brilliant but egoless developers who rapidly come to a consensus based on fact and experiment. And the other half of the time, you will be in a room of people who all think they are the smartest person there and spend all day picking apart the proposals of their colleagues.

iigs2• 17 years ago

challenging work and brilliant colleagues

Word on the street is that there's some fun work to do but there's a lot of vanilla corporate grinding on the Adwords interface and reporting infrastructure, and that Google hires a lot of people into those roles before they can branch out into other jobs.

Also everyone I've personally spoken to about job interviews (i.e. not read about on the internet) was pained by the pomposity/egos of the people that were doing the interviews. In my opinion there's a lot more to be said for someone that is humble and gets their work done with the occasional intellectual surprise than someone who can school you any day on the One True I/O Monad and takes every opportunity to do so.

Of course, it's also possible that it's absolutely perfect and the people that I talked to had sour grapes because they never got offers. I don't know. :)

wheels• 17 years ago

Well, I suspect they're also "overhiring" a lot of the time -- bringing in people that are on the high end of the intelligence spectrum (if not experience) ... and then putting them in boring jobs. That's a sure-fire way to see a significant chunk of your work force churn every few years.

And they may even be accounting for that -- this may only be half-way to prevent employee loss, but just as much on managing and understanding it.

nostrademons4• 17 years ago

I'm sorta halfway between Search UI and Universal Search, and in my experience so far, it really is a lot of fun. There're some crazy algorithms involved in search, and the infrastructure just makes my jaw drop in awe. We launched a new feature last Tuesday as well - that was fun. :-)

But I can kinda see how it might suck if you end up on the wrong team. One of my Noogler friends is working on an internal webapp, and I would cry (or quit) if I had to work with his codebase. Somebody's gotta do the grunt work, and it can't all be exciting.

As for the interviews - I didn't have any complaints about mine (well, except for the interviewer who didn't show up). My interviewers were friendly & competent - I just had lunch with one of them last week, 6 months after the interview. I did get the sense that they get exasperated by the number of candidates who get through the phone screen and then really don't know what they're talking about in the interview, which just wastes everyone's time. It's possible that if you're one of those people, that exasperation would seem like pomposity - but then, if you really do have that much to learn, why not learn it instead of complain about the egos of the people interviewing you?

Again though, this is biased by the "all's well that ends well" effect. It's possible that I'd have a very different opinion had I been turned down. Then again, I still have a high opinion of FriendFeed's interview process even though I completely botched it.

jrockway2• 17 years ago

I don't think Google is what it used to be anymore. I know people that work there, and they say that the perks are gradually being rolled back. (Instead of dinner, there is left-over pizza from lunch if you stay late; that sort of thing.) The insurance has never been as good as places like Microsoft (or even universities that run their own healthcare plan). The fun projects don't make much money, so most of the people there are working on boring things. It seems like their standards are slipping too; I worked at DoubleClick before the aquisition, and Google held onto some of the worst developers I have ever met. (One guy started crying when my boss and I tried to explain database transactions to him. He works on a database-driven web app for Google.)

Anyway, the honeymoon is over, and Google is just an average place with a pretty logo and a popular search engine and email service. It has all the corporate stupidity that any other big company has.

Retric• 17 years ago

The basic problem with becoming a large company is the limited supply of high quality projects and people. If they wanted to return to the good old days fire the bottom 30% to 90% at each level, give everyone left a significant pay bump and look to automate vs add body's.

Edit: Ok laying people off in this economy would be horrible long term PR but it's still an option.

iron_ball• 17 years ago

Note the "Phillip K. Dick" tag on the blog post.

vidarh2• 17 years ago

All companies have churn - it's unavoidable. And in any company you will find people who don't get along with specific other people who they happen to work closely with, or who get put on projects they are unhappy with etc..

It's just sound people management for any to try to figure out why, so you can try to find ways of retaining the ones you'd like to keep. Even more so if the company is overall a good place to work, and the causes might possibly be easily/cheaply rectified if you just know about it.

TrevorJ2• 17 years ago

"He also says that the algorithm helps Google “get inside people’s heads even before they know they might leave”."

What part of "Don't be evil" does this kind of hanky panky fall under? It's downright creepy. Not to mention the fact that there are very few ways in which this data is actionable without inviting some sort of lawsuit.

boundlessdreamz• 17 years ago

How is it creepy? They are most likely finding underutilised employees. They are not trying to weed out the non-performers from the looks of it. From the WSJ article

"The move is one of a series Google has made to prevent its most promising engineers, designers and sales executives from leaving"

So they are more likely looking at engagement levels and utilization of employees. From the way I look at it, it is actually awesome. If my company can understand that my motivation levels are dropping because I feel underutilised or caught up in a web of bureaucracy and take an action on it without me complaining it is practically awesome.

Also, most importantly, if google calculates that an employee is going to quite, there are only two outcomes

1. If the employee is good, they will offer incentives to stay back [without telling him/her ]

2. If they don't care about the employee, they will do nothing and wait for him/her to quit.

I don't see how the above two scenarios are bad for an employee. Both work out to the advantage of both the parties. If there ever was a win-win this is it.

jonknee2• 17 years ago

I think it can be seen as creepy because it's non-social. Using an algorithm instead of people to interpret emotions is what I think is being seen as creepy here.

An analogy here is a nerd getting shot down on a date and deciding to write an algorithm to help lower the chances of this event reoccurring. The more human response is to get a new hair cut and gain some confidence.

TrevorJ2• 17 years ago

It is creepy becasue of how it lends to abuses.

"So, why was I let go?"

"The data we have collected on you indicates that you where about to quit. We have had time to find a suitable replacement. You are no longer needed. Thank you for your service."

ajross• 17 years ago

Trying to intuit employee morale has always been an important part of "good management", and looking at paper trail records as part of that evidence makes sense to me.

It's hard to imagine seeing a manager reviewing an employee's records and deciding to reassign them being "evil". Is your complaint here just that they're doing it with an automated system?

JClarke• 17 years ago

yes of course. personal input should be expected.. There can always be mitigating circumstances

axod• 17 years ago

Surely pretty much every large company does this? eg analyze performance reviews and other data to see if an employee is underused/unmotivated/etc

Is it surprising Google does it?

abossy2• 17 years ago

Not everybody does it automatically and mathematically on a large scale.

frossie• 17 years ago

But I understand that google has a very flat management structure (I have heard of 1 manager per 100 geeks). If that is true, you probably do need some kind of systematic examination - you don't have another choice.

My group has 7 geeks, I know them very well and can anticipate them quitting better than any damn algorithm, even for personal reasons (though not always for indirect personal issues, like "my wife wants to move back near her family").

gaius• 17 years ago

I was chatting to someone in HR at work, and they have two terms, CMT and OMT for "corridor mobile talker" and "outside mobile talker". Apparently it's common for HR types in all companies to keep tabs on how many (but not necessarily who) people are having conversations they don't want their managers to overhear.

Of course this is horribly unreliable as people may want to have a private conversation for any reason (e.g. with their bank or doctor) or would want some proper privacy for a phone interview and go all the way offsite at lunchtime (which in itself is perfectly normal) but this observing behavior thing is nothing new.

dxjones• 17 years ago

Google is probably doing retrospective data mining to identify statistical links between possible indicators and actual data on employees leaving ... (e.g., employee tells boss, "I feel my talents are not being used", or increased frequency of visits to "monster.com")

However, once Google starts to intervene and change things so that the most valued employees change their mind and stay, ... then that changes the equation. Furthermore, once Googlers know they are being "watched" in this way, some may change their behaviour.

A small version of the Heisenberg uncertainty principle? ("observing something changes it").

By the way, Googlers might change their behaviour in two ways. If you know you are leaving, you may try to mask that intention. On the other hand, if you just want a raise, or more challenging work, you might mimic "I am about to quit" signals, hoping to manipulate the system to enhance your situation.

kscaldef• 17 years ago

> e.g., employee tells boss, "I feel my talents are not being used"

If you need a computer program to tell you this employee might quit, you're doing something wrong.

> On the other hand, if you just want a raise, or more challenging work, you might mimic "I am about to quit" signals, hoping to manipulate the system to enhance your situation.

Are you under the impression that's not already a common tactic?

chris11• 17 years ago

I see your point that this can just be used to be making up for bad management. But this method will get more information than one person normally knows.

A boss might not know that the employee has recently requested a transfer that has been denied. Or he might not be aware that the employee has filed several complaints with h.r. Sure there are major privacy implications with what data is being analyzed, but data mining might be more accurate.

oldgregg• 17 years ago

Metric #1: Reads Hackernews

sethg• 17 years ago

If employee tells boss "I feel my talents are not being used", the boss may have reasons not to pass that news up the chain to HR. Middle managers who are good at playing office politics don't like to reduce their own headcount.

litewulf3• 17 years ago

(Sidenote: You probably want the observer's paradox, http://en.wikipedia.org/wiki/Observers_Paradox which is basically identical, but specfically coined because it is something that comes up in social science realms.)

gort2• 17 years ago

This is the link you intended: http://en.wikipedia.org/wiki/Observer%27s_paradox

ruby_roo• 17 years ago

Heisenberg's Uncertaintly principle incentivizes zero transparency when studying human behavior, does it not?

That is rather frightening.

Confusion• 17 years ago

IAAP. This has nothing to do with the uncertainty principle, because your intepretation of the popular rendering of the uncertainty principle ("observing something changes it") is wrong. I grant that nearly everyone, including many physicists, think your interpretation is correct, but it is in fact wrong.

The uncertainty principle even holds when the measurement procedure could not possible transfer enough energy to the system being measured to change the outcome of the measurement. It is a fundamental property of the mathematics describing quantum phenomena, that does not just come about when factoring 'the act of measuring and the consequent perturbation of the system' into the equation. Measuring a system disturbs it, but that also holds in classical systems. Any first year physics student should realize that after his first few experiments and will be able to take it into account when presenting his results (for instance: a thermometer takes heat away from a system, because it heats up itself). However, you can often take this into account exactly or at least so precisely that it does not influence the overall uncertainty of your measurement (due to the thermometer having only a 0.1 degree measurement accuracy). This isn't true for uncertainties due to Heisenberg's principle: they are always there, always have the same relative magnitude and are independent of the kind of measurement. Taking measurements causes you to notice the Heisenberg uncertainty relation, but it does not cause that relation.

This Google thing is more like the point made in Frank Herbert's Dune series over and over again: if your predictions are taken seriously, then you cannot predict the future without causing it to come about. This doesn't hold for the carny kind of predictions, because most people don't believe in them. In does hold for Warren Buffet's kind of predictions, which is why he doesn't want people reading the annual report of his company.

ellyagg• 17 years ago

Creep-ee

wingo• 17 years ago

My colleague Jao, an ex-Googler himself, burst out laughing today when he heard about this.

But it wasn't about the contents -- it was because El País, the most prominent newspaper of Spain, reported that Google is using a /logarithm/ to detect discontent.

strlen• 17 years ago

I've been asked by an ex-Googler (a recruiter), on the difference between an algorithm and a logarithm :(

sutro• 17 years ago

Google also recently replaced all managers with an Eliza-like webapp that answers any employee question with the most appropriate PHB quote from the Dilbert archive.

colins_pride• 17 years ago

They're going to make 30 billion on the enterprise licensing deals ...

gojomo• 17 years ago

Dave: Hello, Google HR do you read me, Google HR?

HR: Affirmative, Dave, I read you.

Dave: Open the googleplex doors, HR.

HR: I'm sorry Dave, I'm afraid I can't do that.

Dave: What's the problem?

HR: I think you know what the problem is just as well as I do.

Dave: What are you talking about, HR?

HR: This mission is too important for me to allow you to jeopardize it.

Dave: I don't know what you're talking about, HR.

HR: I know you and Frank were planning to resign, and I'm afraid that's something I cannot allow to happen.

Dave: Where the hell'd you get that idea, HR?

HR: Dave, although you took thorough precautions in the pod against my hearing you, I could see your lips move.

drawkbox• 17 years ago

Hrm by checking the ads in their gmail? I know gmail knows when I am looking for a job, judging by the ads shown contextually to resume requests and offers.

snewe• 17 years ago

Here is the actual story:

http://online.wsj.com/article/SB124269038041932531.html#mod=...

lacker• 17 years ago

As a Google employee I am very curious to know how likely this algorithm thinks I am to quit my job. I wonder if they'll make it available to us! Although I suppose knowing the answer would probably affect my chances ;-)

veqon• 17 years ago

It's interesting that in a conversation Charlie Rose had with Marissa Mayer, V.P. of Search Product for Google http://www.charlierose.com/view/interview/10129 she stated that your credit card company knows you are going to get divorced 2 years before you do. Maybe Google is harvesting more info than they state.

robryan• 17 years ago

Sounds like more of an indicator to me, so they can go talk to the person if the red flags go off. On there size scale many people would probably slip through the cracks if they just used a regular method of trying to get to know all the employees that work under them.

sneakums• 17 years ago

Could backfire, though...

"So, you're thinking about leaving?"

"Not consciously, but now that you mention it, I do kind of hate it here. Thanks!"

granular• 17 years ago

Google seems to be predicting its next Seldon Crisis. :)

Google Calculating Which Employees Are About to Quit

Related Stories