>An analysis of which factors made employees more likely to quit yielded some surprising results: “Those employees who had been promoted more times were more likely to quit, unless a more significant pay hike had gone along with the promotion,” Mr. Siegel wrote.
Sounds like it's just a simple case of disparity "hey, I know I can get a much higher salary elsewhere given my experience and title. HP hasn't given it to me over the last few years, so it's time to go."
Also, as you get promoted, you acquire additional responsibilities which bring with them additional stress and frequently longer hours. Plus tedious or painful stuff like endless meetings and doing performance reviews and firing people. It can get old rather quickly, so you need to give people lots of money to convince them to keep doing it. If a company needs big data and guys with PhDs to figure this out, they're not paying attention to what's going on around them.
Some people say they'd write code even if they didn't get paid, but I haven't heard anybody say they'd be a manager for free.
I think this is neat, though I wish the article had more details!
At first my reaction was obvious: "Well of course you can tell which employees are going to quit! If they're not getting paid well, don't feel like they have mobility, and don't have flexibility, they will move on!"
But so much of what you read about "what employees want" is both anecdotal and highly-personalized. Some employees want to work from home and others prefer face-time. Some folks want a decent salary, whereas others (like myself) choose something less lucrative in order to work at a startup. For some, upward mobility equates to a track to management, and for others a tech lead, and yet others just want to be left alone to write some darn code.
But if we can try and predict, using actual real data, what kinds of corporate knobs (raises, promotions, job rotations, etc) can decrease attrition, then that also means a company would be increasing job satisfaction in the ways which are important to employees. Why haven't we been doing more of this all along?
Trying to predict that with data and algorithms is bollocks...
There is someone, called manager. I know, I know, it has a bad reputation. But is supposed to talk from time to time and keep an eye about how things are going. If talk is done properly, can be very useful to know if someone is not happy.
I know that we are engineers, we love data, and talking to someone is awkward and out of the comfort zone. But this is ridiculous...
How do we evaluate whether a one-on-one is productive? How can a business, in aggregate, learn from the hundreds of independent weekly manager-employee meetings to see what kinds of issues keep cropping up?
This of course presumes both a good manager and an employee who is willing to speak candidly with them. This may not always be the case. (How often is it not the case? Let's examine the data and see! Maybe there is a systemic problem with trust between managers and their employees that needs work?)
It is not uncommon for businesses to periodically do "employee satisfaction surveys." I don't think crunching data to measure the effectiveness of management (which affects employee attrition) is ridiculous at all.
Just another poor implementation of tech into enterprise really. Why not vet employees rather than chastise them? The top down approach is a failure at all levels.
Very similar analytics work takes places in the telecom industry with respect to customer churn. However, this typically requires a large data size for it to be accurate. So, I wonder how many companies are large enough to benefit from it.
On a larger note, I wish "managers" used other simple means like talking to people openly about their aspirations, instead of blindly following the processes laid out by the company. Thankfully I don't work in such a place - but most big corps could just do better on the attrition rate if they made their appraisal process more employee friendly, focussed on encouraging a open culture and reduced needless hierarchy.
And this is why I'll never work for a company run by MBA driven bull-crap.
Any properly managed business group will have this covered because they treat their human "resources" like actual humans - not variables in a spreadsheet. A GOOD manager would know an employee was at risk as a result of doing their god damn job properly. You do not need a team of data scientists to figure this out.
Advice for younger engineers: if a company has grown past 100 employees, stay away from it. It's inevitably laden down by morons who would forego talking to you on a personal level in favor of a mechanical system with as little interaction as possible (liability and all).
I believe these kind of systems are implemented in much smaller companies as well. The last company I worked at started implementing mechanics to determine roles and payment when the company grew to 40 people or so.
The system that was introduced annoyed me, since it seemed to favor people in a management roles and also with a Masters. The system was created by consultants on behalf of my employer. To me the system felt like me being a horse with a carrot held in front of me. I guess it's true what's being said about engineers and herding cats. Engineers do want a personal treatment.
This rewarding system was one of the reasons I decided to leave later that year. It seems the turnover inside said company has increased shortly after I left, perhaps other people felt the same.
I think the flaw in "MBA driven bull-crap" is an incomplete algorithm. You can only track what's being measured, and only if you understand what's being measured fully. Humans are extremely complicated creatures, so currently they're impossible to measure accurately, and to fully understand. The refined technology simply doesn't exist yet.
Smart MBA's will recognize this and adjust things accordingly, but this has to be embraced throughout the organization. One minnow is easily overlooked in a sea of fish.
--------------------
Personally, I like to measure companies based on how well they implement idioms from other complex professions, like programming. DRY, KISS, etc... The more repetition, the more unneeded complexity, the poorer the company is to work for (IMHO).
If I was going to run such a program (whilst stroking my white cat[1]...), I'd correlate website access to attrition. This could uncover not only active job searching, but general boredom and dissatisfaction surfing patterns...
I'm guessing obvious websites like linkedin would pop up, but I wouldn't be surprised if other patterns would emerge.
[1] Companies, especially of this size, already do this kind of spying on their staff. Not necessarily with cats.
I should probably follow some of the links referenced in the article. The article itself is a little irritating. It seems like the goal is askew: predict who will leave so you can keep them. In every organization I've been in, some people should leave. Perhaps they're at a bad point in their lives; perhaps they're just not a good organizational fit. Whatever the reason, keeping them poisons the team they work with.
The approach is seriously lacking in the factors it looks at for flight-risk. Every organization has a culture: a way of going about getting work done, dealing with petty administrative tasks, behaving, talking, etc. These factors are extremely difficult to convey when hiring someone. Unfortunately, some individuals aren't always a good fit for a culture.
Over time, I've found out from experience that I don't like large organizations. Large organizations have a certain bureaucracy and formality that I find irritating. I left one particular organization just for that reason. My current employer is still too large for my tastes. They treat me well, the pay is fine, the benefits are unheard of nowadays. I just don't like the bureaucracy and formality.
One more point is that the reasons for leaving may be different in our current era of entrepreneurship. People may simply leave to follow their own ideas about a business.
This is a sales pitch by HP - they are trying to sell 'big data analytics' to huge clients, like General Electric. Oracle and IBM are also trying to sell this stuff. It's supposed to be the next big thing.
I think there is some doubt here that it is valuable, which fails to appreciate scale.
If someone said to you 'I don't know why Twitter need all these engineers - I made a site like twitter in 2 hours in an introduction to rails tutorial', you'd explain that scale was the problem.
The CEO of Toshiba oversees business units that do nuclear power plants, LCD screens, and business systems. His understanding of any of these units is fairly shallow - he knows less about nuclear power plants than the CEO of that unit, who knows less than his chief engineer. The chief engineer knows less about the workstation login page than the junior who implemented it.
When you are working at that scale, your tools and information are limited and broad. You can set budgets, issue directives, and try to create culture.
On the other hand, because you are working with large numbers of people (data points), it makes sense to make decisions statistically.
When N is 1, you need to make decisions subjectively - it doesn't matter that taking Statins after 50 slightly reduces the chance of heart disease for some people (possibly including you). You decide if you want to take a pill every day, and broadly if you want to do things that are healthy vs things that are enjoyable.
For the guy deciding whether to spend millions on Statins as part of your health insurance, it makes total sense - he can model how much heart disease costs, and how much statins reduce that, along with some value placed on an extra year of life, plus something about cost to acquire customers and policy retention, and decide if it's worth the money. So the British NHS decided to give everyone over 50 Statins.
The guy in charge of HP can model how much staff churn costs, and assign budgets that are sub divided down by business units. Using this data he can target that budget where it will help the most.
And that is all he can do.
Another thing is that it doesn't matter if it's a good idea, it matters if someone will pay for it. If HP can demonstrate measurable financial advantage to doing this they can land clients. That's a lot easier to measure = sell than 'not being a douche' training for managers.
27 comments
[ 4.8 ms ] story [ 66.9 ms ] threadSounds like it's just a simple case of disparity "hey, I know I can get a much higher salary elsewhere given my experience and title. HP hasn't given it to me over the last few years, so it's time to go."
Some people say they'd write code even if they didn't get paid, but I haven't heard anybody say they'd be a manager for free.
At first my reaction was obvious: "Well of course you can tell which employees are going to quit! If they're not getting paid well, don't feel like they have mobility, and don't have flexibility, they will move on!"
But so much of what you read about "what employees want" is both anecdotal and highly-personalized. Some employees want to work from home and others prefer face-time. Some folks want a decent salary, whereas others (like myself) choose something less lucrative in order to work at a startup. For some, upward mobility equates to a track to management, and for others a tech lead, and yet others just want to be left alone to write some darn code.
But if we can try and predict, using actual real data, what kinds of corporate knobs (raises, promotions, job rotations, etc) can decrease attrition, then that also means a company would be increasing job satisfaction in the ways which are important to employees. Why haven't we been doing more of this all along?
There is someone, called manager. I know, I know, it has a bad reputation. But is supposed to talk from time to time and keep an eye about how things are going. If talk is done properly, can be very useful to know if someone is not happy.
I know that we are engineers, we love data, and talking to someone is awkward and out of the comfort zone. But this is ridiculous...
This of course presumes both a good manager and an employee who is willing to speak candidly with them. This may not always be the case. (How often is it not the case? Let's examine the data and see! Maybe there is a systemic problem with trust between managers and their employees that needs work?)
It is not uncommon for businesses to periodically do "employee satisfaction surveys." I don't think crunching data to measure the effectiveness of management (which affects employee attrition) is ridiculous at all.
Analysts: See, they're no longer employed! Another successful prediction for the algorithm!
https://news.ycombinator.com/item?id=617533
On a larger note, I wish "managers" used other simple means like talking to people openly about their aspirations, instead of blindly following the processes laid out by the company. Thankfully I don't work in such a place - but most big corps could just do better on the attrition rate if they made their appraisal process more employee friendly, focussed on encouraging a open culture and reduced needless hierarchy.
Any properly managed business group will have this covered because they treat their human "resources" like actual humans - not variables in a spreadsheet. A GOOD manager would know an employee was at risk as a result of doing their god damn job properly. You do not need a team of data scientists to figure this out.
Advice for younger engineers: if a company has grown past 100 employees, stay away from it. It's inevitably laden down by morons who would forego talking to you on a personal level in favor of a mechanical system with as little interaction as possible (liability and all).
The system that was introduced annoyed me, since it seemed to favor people in a management roles and also with a Masters. The system was created by consultants on behalf of my employer. To me the system felt like me being a horse with a carrot held in front of me. I guess it's true what's being said about engineers and herding cats. Engineers do want a personal treatment.
This rewarding system was one of the reasons I decided to leave later that year. It seems the turnover inside said company has increased shortly after I left, perhaps other people felt the same.
Smart MBA's will recognize this and adjust things accordingly, but this has to be embraced throughout the organization. One minnow is easily overlooked in a sea of fish.
--------------------
Personally, I like to measure companies based on how well they implement idioms from other complex professions, like programming. DRY, KISS, etc... The more repetition, the more unneeded complexity, the poorer the company is to work for (IMHO).
I'm guessing obvious websites like linkedin would pop up, but I wouldn't be surprised if other patterns would emerge.
[1] Companies, especially of this size, already do this kind of spying on their staff. Not necessarily with cats.
This is only a preliminary result though, we need more funding to back up this radical hypothesis.
The approach is seriously lacking in the factors it looks at for flight-risk. Every organization has a culture: a way of going about getting work done, dealing with petty administrative tasks, behaving, talking, etc. These factors are extremely difficult to convey when hiring someone. Unfortunately, some individuals aren't always a good fit for a culture.
Over time, I've found out from experience that I don't like large organizations. Large organizations have a certain bureaucracy and formality that I find irritating. I left one particular organization just for that reason. My current employer is still too large for my tastes. They treat me well, the pay is fine, the benefits are unheard of nowadays. I just don't like the bureaucracy and formality.
One more point is that the reasons for leaving may be different in our current era of entrepreneurship. People may simply leave to follow their own ideas about a business.
I think there is some doubt here that it is valuable, which fails to appreciate scale.
If someone said to you 'I don't know why Twitter need all these engineers - I made a site like twitter in 2 hours in an introduction to rails tutorial', you'd explain that scale was the problem.
The CEO of Toshiba oversees business units that do nuclear power plants, LCD screens, and business systems. His understanding of any of these units is fairly shallow - he knows less about nuclear power plants than the CEO of that unit, who knows less than his chief engineer. The chief engineer knows less about the workstation login page than the junior who implemented it.
When you are working at that scale, your tools and information are limited and broad. You can set budgets, issue directives, and try to create culture.
On the other hand, because you are working with large numbers of people (data points), it makes sense to make decisions statistically.
When N is 1, you need to make decisions subjectively - it doesn't matter that taking Statins after 50 slightly reduces the chance of heart disease for some people (possibly including you). You decide if you want to take a pill every day, and broadly if you want to do things that are healthy vs things that are enjoyable.
For the guy deciding whether to spend millions on Statins as part of your health insurance, it makes total sense - he can model how much heart disease costs, and how much statins reduce that, along with some value placed on an extra year of life, plus something about cost to acquire customers and policy retention, and decide if it's worth the money. So the British NHS decided to give everyone over 50 Statins.
The guy in charge of HP can model how much staff churn costs, and assign budgets that are sub divided down by business units. Using this data he can target that budget where it will help the most.
And that is all he can do.
Another thing is that it doesn't matter if it's a good idea, it matters if someone will pay for it. If HP can demonstrate measurable financial advantage to doing this they can land clients. That's a lot easier to measure = sell than 'not being a douche' training for managers.