Mathematical Modeling Used To Track and Label

Anti-Globalism writes to tell us that in a new book titled The Numerati, author Stephen Baker introduces us to some of the math wizardry that is used to label or track our movements through purchases, phone calls, internet usage and other habits. "One of the most promising laboratories for the Numerati is the workplace, where every keystroke, click, and e-mail can be studied. In a chapter called "The Worker," Baker travels to IBM, where mathematicians are building predictive models of their own colleagues. An excerpt: 'Samer Takriti, a Syrian-born mathematician. He heads up a team that's piecing together mathematical models of 50,000 of IBM's tech consultants. The idea is to pile up inventories of all of their skills and then to calculate, mathematically, how best to deploy them. I'm here to find out how Takriti and his colleagues go about turning IBM's workers into numbers. If this works, his team plans to apply these models to other companies and to automate much of what we now call management.'"

Comments

pseudo-continuance

[easyTree]

If this works, his team plans to apply these models to other companies and to automate much of what we now call management...

...The management function is reputed to make heavy use of the functions 'rand'.

Re:pseudo-continuance

[William Robinson]

Look at positive side. Finally, I would get manager, who knows 2 plus 2 = 4.

Re:pseudo-continuance

[aadvancedGIR]

When I was a consultant, one of my mangers was more a garbage collector: he tagged me "CORBA expert" and was ready to sell me on that ground after I added a new function (mostly by copy-paste) in an object's IDL.

Not a chance.

[aadvancedGIR]

If managemenet was really about optimizing resources, it would have been outsourced a long time ago.

As a non-manager, I can tell you the most important job of management is to deal with the unquantifiable: engineers need to feel unique and usefull and they need opportunities to work on new things (and/or be promoted) from time to time. A good manager knows his guys are much more than their previous experiences (and somtimes slightly less too).

Re:Not a chance.

[slim]

All true. But parts of IBM (and I'm sure the rest of the corporate world) have already forgotten that.

I used to work there, and there was a big effort on to have employees maintain a 'skills' database. It was clear that despite running top class courses on teams, there were influential people in the corporation who saw staff as being nothing more than a set of D&D type stats who could be deployed like pawns.

And hey, the losses in morale, effectiveness and customer satisfaction might be offset by the cost reductions. Who knows. I'm just glad I don't work there any more. (Which is quite lucky, because I didn't resign - my business unit was sold).

Re:Not a chance.

[zunicron]

We can get computers to make us feel special, no worries.

Re:Not a chance.

[Mark_in_Brazil]

I've never worked at IBM, but I've known people who did, including my ex-wife, back when we were married.
I'm not sure about elsewhere, but here in Brazil, IBM attracts employees. One is by being multinational tech giant. People who value stability and like to say they're in a big company (there's a lot more of that than I ever would have imagined) are attracted by that image. Back in 2000, when my then-wife was at IBM, I knew one person (a woman, but not my then-wife) who got PMP certification and had done a lot of training at IBM, and was getting a lot of attention from headhunters. She was given the opportunity to interview for a job with twice (TWICE) the salary of the job she then had at IBM, but didn't even try to find out more about the company or go to the interview because, in her words, "I've already got a nice little career at IBM, so I'm going to stay." My first thought was that IBM, like most other publicly traded companies, would "downsize" by purging a four- or five-digit number of jobs, and would do it without blinking. That is, IBM would be nowhere near as loyal to this person and thousands like her as she was being to IBM. The thing is that I realized she believed her job was safe because the company is big, and if I had said what I was thinking, we would end up in an argument how well her job at IBM might weather tough times, and her image of IBM's stability was much too deeply rooted for me to change it.
The other thing about IBM that attracted people to work there is that IBM was known for giving its employees lots of training. Here in Brazil, a lot of tech people I met made frequent mentions of "Faculdades IBM" (roughly, "IBM University"). It was a place you went and earned a salary while learning new skills and new technologies free. Yes, the salary was less than you could earn at another job, but the training made it worthwhile, because after a few years at IBM, you could find a much higher-paying job with your new skills and experience. IBM was kinda screwing up by letting its employees get away, and that was largely because annual salary adjustments for loyal employees were small enough that even some of the stability-seekers were tempted to look elsewhere.
When I was back in the US for the last time before moving to Brazil, which means somewhere between April and June of 2000, I met a friend of friends who was working at IBM somewhere in California. I told him about the "IBM University" image the company had in the Brazilian high-tech market. He told me it was similar in the US. I mention that it was only one person, because this may not be generally true, but in the view of this one friend of my friends, it was. In fact, he told me he was earning a lot less than similarly-qualified friends, and some had even tried to get him to go and work with them, but he had a multi-year plan involving lots of training and experience at IBM before hitting the job market. He wanted to have a resume with training and experience that would get him the job he wanted without the job-hopping approach his friends were taking. Again, this was what one IBM employee told me in 2000, so I don't want to generalize.

All the problems mentioned in the parent post, plus some real jerks who were managers, plus some really ridiculous rules imposed on employees (gawd, the e-mails parodying those rules were hilarious, and I knew enough people at IBM Brasil to get several copies of each), contributed to IBM Brasil's less-than-ideal work environment. But IBM was able to keep recruiting even good employees because the employees, for one reason or another, believed it was worth dealing with that. The people whose thinking was stuck in Brazil's more unstable economic past valued the perceived stability of having a job at IBM and being able to proudly tell people they worked at an enormous company enough to deal with the negative aspects of working there. I'm the opposite of these "corporate size queens;" I never liked working at a company larger than a given size. Given

Re:Not a chance.

[MurphyZero]

I switch jobs about every two years. I too enjoy the ability to learn something new and/or get out of something boring.

Re:Not a chance.

[Phat_Tony]

Management is largely about optimizing resources, but of the tradeoffs management must consider, only some of them can be quantified numerically at all. The decisions managers have to make are often that of weighing something easily quantifiable and incredibly precise against something hopelessly vague and unquantifiable, which a computer will have no chance whatsoever of grappling with until we have strong AI. There's a pervasive trend in management to put undue weight on the quantifiable aspects of business, when it's a common fallacy to beleive something should be weighted more highly simply because it's quantifiable.

Here's just one tiny example of the kind of decision management has to weigh: setting service-level goals for a call center. Most call centers measure and target a service level, defined as answering [x]% of their incoming calls in [y] seconds or less. There's a tremendous amount of exceedingly tricky mathematical modeling that can be done to determine how to staff best to meet service level efficiently, and super-advanced computer programs could play a huge roll in improving this. And good computer modeling can create a beautiful and accurate graph of exactly how much it costs the company to run different service levels. But the key management problem here is setting the target service level, and while you can quantify, model, and analyze the costs of providing that service level intricately, and smart management can optimize that out the wazoo and bring ever lower costs to providing the same service level, it's almost impossible to gather the tiniest shred of evidence regarding the benefits of different service levels.

Sure, we all know what the benefits are- how many people get sick of waiting and call another company? How many do so subconsciously? How many only do so after years? How many of your customers tried another company once, and made the connection that that company provides poor customer service because the wait on the phone was so much longer, or switched because it was so much shorter, or the opposite- competitor's customers who did or did not switch to you because of the same? How much money did you save in returns or less complaint calls because you had built up goodwill by always answering the phone fast and not keeping your customers waiting? How about trying to quantify the mental health benefit to your own sales force from having happier, less irate customers, because they weren't kept on hold interminably before you answered? You have no hope of quantifying the benefits, but you must set some service level, based upon your intricate analysis of costs and NO IDEA what the benefits are.

And almost every aspect of business if FULL of decisions like this.

I used to manage a call center, and we answered 97% of calls in 18 seconds or less- that's three rings. There was no computer answering system, and no queue except the ringing. When we went to call-center industry conventions, people literally wouldn't believe that any major call center ran service levels that high, and if they believed it, they'd tell us we were insane, that that service level represents an unconscionable waste of resources. For comparison, many companies had targets more like 60% of calls in 10 minutes or less, with a computer holding queue. Some government departments (I'm not kidding) had goals of 50% of calls answered period- that is, answered before the caller bailed out of the holding queue by giving up, with no time factor.

We strongly disagreed that we were significantly overspending on service level, but there wasn't anything to even say to argue about it- the value of a good customer service experience is just ridiculously difficult to quantify. But even at that company, I routinely saw people making the kind of mistake we felt companies who kept their customers on hold forever were making- they tend to move in bias of the information they can quantify. Some manager goes to a meeting of higher-ups and says "I found out how we can save $6 million a yea

So scary its..... yawn.

[Eth1csGrad1ent]

Yawn. Take a look at my life if you want, but let me save you the calculations. I had a busy, if somewhat mundane, day at work after which I came home, ate some pizza, caught an episode of Top Gear and Boston Legal, checked in on my facebook account, and - as the missus is away - there is probably some internet porn in my near future. Tomorrow, being a Tuesday, will most likely turn out to be a carbon copy of today, with the following exceptions, Top Gear becomes Criminal Minds followed by NCIS instead of Boston Legal. THE END.

Re:So scary its..... yawn.

[rasputin465]

Well, I generally come in at least fifteen minutes late, ah, I use the side door - that way Lumbergh can't see me, heh heh - and, uh, after that I just sorta space out for about an hour. I just stare at my desk; but it looks like I'm working. I do that for probably another hour after lunch, too. I'd say in a given week I probably only do about fifteen minutes of real, actual, work.

Management now?

[houghi]

automate much of what we now call management.

That means I will be managed by a mindless droid. Not much difference from now and most likely even an improvement from the humanoid random generator that takes decisions now.

Re:Management now?

[aadvancedGIR]

The good thing is that if you spend your entire day on economy and stock exchange sites, you might trick the system into promoting you to CEO position, of course, you need avoid /. and revert back to unpatched IE6 too.

an excellent way to reduce the quality of workers

[thermian]

Actually, a lot of consultants are highly skilled people who do not have to work for any one person.

Automate their management, and you'll start making them feel like factory workers. Smart people are far less likely to accept inflexible working conditions. The result will be that they walk.

I know I would. My consultancy work is expensive, and I insist on doing what I want, when I want, for who I want. Ok, I'm picky, but I'm happy and I enjoy what I do, so the quality of my work remains high.

If someone started dictating things I had to do based on a mathematical model, I'd go elsewhere for a more relaxed environment.

End of nepotism, at last!

[aadvancedGIR]

Good thing: the manager doesn't have a stupid son or nephew to promote instead of you.

Bad thing: no more open position available for promotion. (and so the death of the compagnies caused by the unability to apply Dilbert's rule).

Similar concept has been tried before

[francium de neobie]

By Frederick Winslow Taylor, who pioneered the concept of scientific management.

While scientific management has its uses (e.g. optimizing an industrial process), it is definitely not everything about management. Scientific management has received plenty of criticisms in the past when it's overused, especially when the manager pushed it to the extreme and overspecialized the roles of employees - it devastates morale and harms everyone in the long term.

So while I respect the work these guys are doing, the "to automate much of what we call management" bit sounds like an exaggeration to me. After all, a lot of management is about people and communication, and even our best AIs don't have much idea about the latter today.

Mathematics Is A Harsh Mistress

[ObsessiveMathsFreak]

You need to be careful about applying mathematics. Typically a mathematical model will have assumptions from which predictions can be made. The trouble is the instant you make your assumptions, the predictions become locked in. They are predetermined, even though you may not have discovered them yet. And of course if your assumptions are wrong, or inaccurate, your predictions are not going to match the real world very well. You can't massage the data or your results to get around this. Once your assumptions are made, mathematics leave no room for debate. Ever.

What this means of course is that it is often complete folly to apply mathematics to complex human interactions. Any assumptions you make will be totally inadequate to fully encompass any large organization and its members, and as a result, your predictions will probably be erroneous. Proceeding to apply your derived results to people will lead to unsatisfactory results and unexpected effects.

The Adam Curtis documentary The Trap, discusses the problems in reducing industries, outputs and people to numbers. Basically the numbers, which are more or less wrong, force people to conform to them, and you end up breaking existing systems completely or else converting them into an inefficient version of themselves, all the while thinking (and being told by the numbers) that your systems are improving.

The power of a mathematical model and its caveats, are in fact best described by Douglas Adams' fictional supercomputer Deep Thought. Deep Thought could indeed provide the answer to Life the Universe and Everything(42), and it was perfectly correct. But it wasn't any good because no one knew the correct Question to ask in the first place.

Mathematics will give you answers, and they will be right. But you had better be sure that you asked the right questions before you act on them.

Re:Mathematics Is A Harsh Mistress

[plopez]

What about stochastic methods?

OMG - Protect your job ASAP!

[Linker3000]

Quick, start doing something random, but work-related, regularly, at random intervals.

Model that you bastards.

Re:OMG - Protect your job ASAP!

[Linker3000]

Forgot to add: This reminds me of a cartoon I saw perhaps 10-15 years ago; a potential employee is being shown around by a Manager. The caption read something like:

"Of course, we like to treat everyone here as individuals - for example, this is individual #64881"

Qualitative vs Quantitative

[gusmao]

According to the article, they don't have access to evaluation performance reports, so they are using basically hard data (as salary, programming languages used, experience in particular projects, etc) to model the very broad concept of "skill".

The problem is that the evaluation reports are precisely the most likely source to tell a good programmer from a just regular one, which is exactly what they don't take into account. A junior programmer may be excepional, but he will have less experience and a much lower salary than a "senior" programmer that could just market itself very well, and therefore be marked as a commodity

Categorizing programmers and treating them like commodities could be very dangerous if your model is flawed or considers the wrong parameters, especially if it provides a excuse for bad managers not to think or evaluate their employees individually (hey, if the model says it can't be wrong, right?)

Re:Tautology?

[pjt33]

If you're not "modelling with maths" then you're modelling with something else (astrology? guesswork? religon?)

Clay?

By coincidence...

[pjt33]

Did you notice that the /. fortune at the moment is

If the facts don't fit the theory, change the facts. -- Albert Einstein

?