Game Theory in Management

In my blog from two weeks ago, I promised to present a way to advance a given capability (in this case, project management) within an organization following the structure inherent in Carnegie Melon University’s Software Engineering Institute’s (SEI’s) Capability Maturity Model® while avoiding the dreaded “slow roll” (Professor Bud Baker’s term) or silent veto, and that this method had its roots in game theory. But then, in last week’s blog, I took advantage of the occasion of Kristen Wiig hosting Saturday Night Live to lambast their writers for being highly formulaic in their sketch-writing, so it falls to this, the last week of May (and May’s theme) to provide this answer. And the answer is…

…rooted in Game Theory, like I said before, specifically the game known as the Prisoner’s Dilemma. There are various versions of the Prisoner’s Dilemma, the most common being that you are in a situation where you are in prison, and your jailer tells you that if you rat out (inform on) your cell-mate, he will reduce your sentence. Problem is, you know your cell mate will receive the same offer. There are four possible outcomes:

·         You both rat on each other, and both serve 11 years.

·         Neither of you rats on the other, and you both serve 18 months.

·         You do not inform on your cell-mate, but he informs on you. You serve 18 years, and your cell-mate walks free (this is known as the “sucker’s payoff”).

·         Similarly, you inform on your cell-mate, but he does not inform on you. You walk, and your cell-mate gets 18 years.

Conventional wisdom on the Prisoner’s Dilemma had the best tactic as always ratting out your cell-mate in order to avoid being on the receiving end of the Sucker’s Payoff.

In the 1980s, Robert Axelrod hosted a tournament for competing computer programs playing 200 iterations of the Prisoner’s Dilemma. As  you can imagine, programs employing all manner of strategies were entered, some, keeping with traditional wisdom, informing on the other program every time, others ratting out the first fifty times, but not the next one hundred and fifty, and so forth. One program, submitted by Anatol Rapaport and called Tit for Tat, won the tournament handily. The Tit for Tat program employed the following strategy:

·         The first iteration, it cooperated (did not inform).

·         Thereafter, it did whatever the other program did in the previous iteration.

After Tit for Tat won the overall tournament, some analysts tinkered with its underlying strategy, and developed versions that, say, cooperated the first five times, and then did the tit for tat thing, or else started off by ratting, and then going into the tit for tat strategy, but the alternates always lost to the original. So, the analysis yielded three primary reasons why Tit for Tat won the tournament:

·         It was initially nice – it cooperated the first iteration.

·         It retaliated immediately for defection.

·         It forgave completely and immediately for cooperation.

Many years after I had read the original analysis, and was trying to crack the nut of how to get past people in the organization who were opposed to the advancement of a particular capability by slow-rolling or silent veto, the solution hit me like a thunderclap. In the advancement of a capability (again, in this case, PM) within the macro organization, you really aren’t trying to get everybody to “do” project management: you’re really attempting to get them to cooperate, or at least not defect in the technical program that you are pursuing. It then falls to reason that, in order to secure that cooperation, the implementation phase of the capability maturation program should act like Tit for Tat:

·         It has to be falling-off-a-log easy to plug-and-play.

·         “Retaliate” is a bit strong, but you must be able to identify and respond immediately whenever anyone directly associated with participating in the implementation effort attempts to defect.

·         If the participants are cooperating, then they are golden, even if they are sending you schlock data. Better data can be taught – cooperation can’t.

Following these three steps won’t guarantee a successful implementation, but they will maximize your odds of one. Of course, there are many other factors that can derail your efforts, but I believe having an overall strategy that incorporates these three will improve your chances significantly.

The use of structure in the technical approaches we select in addressing managerial problems allows us to solve recurrent problems with known successful strategies. Often these structures will accommodate analogous situations as well, and we can engage them with a similar expectation of success. But what happens if the new problem isn’t all that analogous, or if the structure is unsound in the first place?

The NBC variety show Saturday Night Live first aired in 1975, and is one of America’s longest-running television shows. It has met with considerable critical and financial success in that time, owing mostly to its mix of talented writers and performers. One week ago Saturday, on May 11, SNL was hosted by former cast member Kristen Wiig, who performed in several sketches as characters she had played during her time as a regular cast member. Her performance reminded me of the many ways that SNL’s tendency to be highly formulaic in their sketch writing has made much of the show, well, unwatchable. Here are some examples of what I’m talking about:

·         Any and all “Gilly” skits. “Gilly” is a mischievous little girl, played by Wiig. She does silly and destructive things, after which her father gravely intones “Gilly!”, and she replies “Yeth.” And, “yeth,” that’s the extent of these skits’ humor.

·         A staple of athlete guest-host, the skit where a sports team is in the locker room at halftime, and dialogue establishes that they are significantly behind. The guest host/team leader tells the story of a previous team that rallied under similar circumstances, having been inspired by some dopey song, a recording of which is played (and usually danced to). In my opinion, this template is vacuous and stupid.

·         Tina Fey’s “impersonations” of Sarah Palin, usually on the Weekend Update segment. These involve no jokes or traditional humor, with the New York City audience howling with derisive laughter at what are essentially nothing more than a string of mean-spirited mocks and insults. When the SNL writers couldn’t mock Palin for anything she had actually said, they made up something for her to say (“I can see Russia from my house!”), and then mocked her for that.

·         “Garth” and “Ann,” supposed folk-music recording artists who readily offer to sing on Weekend Update, but don’t have any songs at the ready. Instead, Fred Armisted and Kristin Wiig sit there and make up songs on-the-fly, while trying to sing the same words and melody. It was never funny, but now it’s foursquare in the Unwatchable category.

·         The skits where a new girlfriend is brought home – usually by Fred Armisted’s character – to meet his extended family, who all greet each other by excessive kissing, usually open-mouthed, even among brothers. Any sign of discomfort at this bizarre behavior comes as a shock to this family. Again, started out as simply unfunny, and through repetition has become unwatchable.

·         “Pig,” the Target cashier, also played by Kristen Wiig. Pig comments excessively on the items that customers purchase, and will go and fetch things for herself when they strike her fancy. Early versions would have a manager step in and announce “Classic Pig!” when Wiig would proclaim some improbable use for various items. Male guest-hosts are invariably shoe-horned in to the skit as stock boys with an unrequited crush on Pig.

·         Film shorts featuring Laser Cats. Cute the first time, not so much the second, and now unfunny and careening into unwatchable, no matter which celebrity is snuck in to the film short.

·         Lady Gaga’s film short on being in a threesome with Andy Samberg and another male cast member (I can’t remember which, ‘cuz it was unwatchable).

·         The “Secret Word” skits. A take-off of Password, Wiig plays the “celebrity” member of one team, but always says the word that her partner is supposed to guess. This “celebrity” is a washed-up Broadway actress who, when not blowing the secret word, waxes interminably about some previous part she had played. Started cute, but through repetition has become unwatchable.

Since Kristen Wiig appears in many of these examples, I’m tempted to reverse-engineer the structure that SNL writers employ and describe it as being characterized by a certain laziness of thought, with an expectation that Wiig’s timing and comic ability will make up for it. I don’t know – maybe there’s even a skit template library that the writers invoke when they are fresh out of fresh ideas. The point here is that, even in the most successful of franchises, structures and templates and canned strategies, even the ones that had perhaps lead to previous successes, can end up resulting in very poor management decisions.

Even on Saturday Night Live.

Building on last week’s blog, where I discussed the insightful framework posited by Carnegie Melon University’s Software Engineering Institute’s Capability Maturity Model®, I pointed out that the CMM®, while asserting the five levels that organizations go through when attempting to advance a given capability, didn’t do as good a job of showing how the organization advances from level to level. One book on this topic by Kim Caputo, CMM Implementation Guide; Implementing Software Process Improvement (Addison-Wesely Professional, 1998) discussed six steps that should take place in-between each CMM® level:

1.      You introduce the nature of the capability improvement to the macro organization,

2.      and then scope out the exact nature of the participant’s activities with respect to the capability advancement.

3.      Launch a pilot project, and (presumably) bring it in on-time, on-budget.

4.      Widespread group assimilation follows the pilot project,

5.      then institutionalization takes place,

6.      followed by an audit to see if you really have achieved the next level.

Now, Ms. Caputo readily admits that there is, in her words, a “chasm” in-between steps 3 and 4, with devastating results. If widespread group assimilation does not take place following the successful completion of the pilot project, the implementation team is forced to start over again, and re-introduce the capability being sought, re-scope out requirements for the actual participants, etcetera, etcetera. But here’s the catch – the implementation team only has two, maybe three bites at this apple before the organization just gets tired of hearing about the need for the particular capability enhancement, and tunes you out.

If I understand Ms. Caputo’s points correctly (and I believe I do), she councils enhanced communications among the implementation team members and those who decline to participate. By drawing out their unstated or poorly-stated reservations, she believes that the parts of the organization that are attempting to avoid contributing to the implementation effort can be persuaded to change their minds, and contribute. Frankly, I don’t agree – I’ve always felt the communications aficionados wildly overstated their techniques’ abilities to improve management.  It was during this time, when I was attempting to solve the problem of members of the macro organization electing not to engage to the extent needed for a successful implementation (a behavior Ms. Caputo calls “the silent veto”), that I had a key conversation with Bud Baker.

Bud Baker is a professor and the Chair of the Department of Management and International Business at Wright State University. However, before he got his high-falutin’ title, he was, like me, a columnist for PMNetwork magazine. Those were PMNetwork’s halcyon days, when the monthly columnist lineup included people like Professor Baker, Neil Whitten, John Sullivan, Paul Dinsmore, Deborah Bigelow, and me (needless to say – but I’ll say it anyway – PMNetwork’s columnist lineup is nowhere near this level of talent today, except maybe when Bud sends in some of his work). While wrestling with this crossing-the-gap issue, I spoke with Bud, who said the act of withholding the participation level needed to perform a given implementation was called “the slow roll” from his days of doing project management consulting for the United States Department of Defense. The organization targeted for the enhanced capability knew that there was a finite amount of energy behind the push for improvement and, if they could participate just enough to see the overall effort lose steam, they would be able to go on doing their jobs the way they had been, without appearing to be the reason the sought-after improvement effort failed. It was both insidious and perfectly predictable.

So, what is the solution to overcoming the dreaded Slow-Roll? The answer comes from a highly analogous situation from Game Theory, and is…

…coming next week (yeah, I know this is the second week in a row I finished with a who-shot-J.R.-style cliff-hanger, but it is a rather complex problem).

There is absolutely no truth to the rumor that I’m paying Cameron under the table to proscribe our monthly themes based on subjects that I’ve extensively researched while writing my books.

None whatsoever.

However, I am, once again, thrilled at the selection of Methods and Frameworks as May’s theme, because my first book, Things Your PMO Is Doing Wrong (PMI Publishing, 2008), was largely predicated on a project management adaptation of Watts Humphries’ seminal book in software management framework, Managing the Software Process (Addison-Wesley Professional, 1989). In it, Humphries introduced the concept of the Capability Maturity Model®, upon which Carnegie Melon University’s Software Engineering Institute® would base its further research on the subject. (Incidentally, in my opinion, PMI®’s attempt at adapting the CMM® to a project management environment, the vaunted Organizational Project Management Maturity Model®, or OPM3®, was just dopey. I was actually in attendance at a PMI® Congress where a member of the OPM3® committee was selling the inchoate mess to the masses. He actually said “We’re developing a structure that will allow people to go where they want to go.” And that was the intellectually insightful highlight of his remarks!)

The Capability Maturity Model®, or CMM®, posits that organizations seeking to advance a capability typically go through five “levels”:

·         Level 1 is represented by chaos. Everybody is doing their own thing, or nothing at all, so that, paradoxically, you can actually have areas of advanced expertise in a Level 1 organization.

·         Level 2 is very basic, but at least everybody is doing the same thing. Same forms, same process, etc.

·         Level 3 is known as “defined.” If the proverbial beer truck hits the heroes who got you to this point, then the capability advances your organization has enjoyed don’t unravel. The written procedures and training are in place to ensure that the macro organization’s capability doesn’t reverse itself.

·         Level 4, known as “managed,” is characterized by your organization being so darned good at the capability that you are in a position to export it to other organizations.

·         Level 5 – “optimizing” – nobody ever really gets to Level 5. Here, your organization is so good at the capability that you are routinely discovering long-standing problems in the industry.

Originally the CMM® was directed at the development of software projects, but other management science specialties and arenas took a look, and basically said “Hey! That’s exactly what happens when we’re trying to advance a capability in our areas!” And, before  you knew it, Carnegie Melon University’s Software Engineering Institute had patented, copyrighted, and just generally legally protected the daylights out of the idea (which is why I’m having to constantly insert the little ® symbol).

Now, the problem with the Capability Maturity Model®, brilliant as it is, is that it doesn’t actually tell you how to advance from Level® to Level® (think I’m being excessive? SEI actually made the word “level” a protected term in the model.). This is where I focused my research for the first book: assuming that the Capability Maturity Model® was applicable to project management, how does one actually advance from Level to Level? I discovered the answer quite by accident, hiding in the realm of Game Theory – specifically, the game entitled The Prisoner’s Dilemma, and its optimal strategy for resolution, Tit for Tat. The way it works is…

Oops! Out of room for this week. I’ll dive in deeper next week, but for the full analysis (and the surprising implications that stem from it), you should probably buy either of my books – or make comments on this blog page along the lines of “hurry up and get us the rest of the 411 on this stuff!”

When April began, I was thrilled that Cameron had given us the theme of Governance. There’s so much misinformation out there that, being the INTJ that I am, I was in the epistemological equivalent of a child in a candy store (or, perhaps more fittingly, a professional skeet shooter in an arcade duck-shoot game). But shooting down invalid or inchoate ideas on Governance is easy – what’s a bit tougher is articulating what proper Governance is, and why.

Not to be gobsmackingly obvious about things, but the true purpose of Governance is this:

Advanced Governance is realized when those in positions of authority make the best possible decisions on behalf of the macro organization.

Ahh, but that’s the rub, is it not? You touch the hot burner, and know instantly that that was a bad decision. You hire the wrong person to head up, say, your PMO, and you won’t know if that was the wrong decision for weeks, months… or ever. How in the world can the manager know if his business decisions are right or wrong before she makes them?

As I discuss in my recently-released, must-have book, Game Theory in Management, these decisions are actually model-able. But the model for these decisions’ evaluation must take into account the fact that there are three different types of management, with different tools and goals:

·         Asset Management deals with the organization’s resources. Its main information source is the general ledger.

·         Project Management addresses how the organization meets its customers’ needs. Its main information source is the Earned Value Management System.

·         Strategic Management deals with how the organization is performing when compared to its competitors. Its main information source is the data feed on market share.

What the manager thrust into the role of Governance must realize is that there are advocates from each of these three orientations on management who will insist that their take – and their take alone – is the most appropriate information feed for upper-level decision-makers, and will expend considerable energy to sell that narrative. Often these management “science” advocates are utterly unaware of where their pet theory exists in the overall scheme of management information streams, but the manager functioning in the role of Governance must know where they are coming from, and the limits to their advocacies.

The manager perfect in Governance will know how each and every decision impacting the macro organization will affect the organization’s standing with respect to its assets, customers, and competitors. However, the macro economy being the near chaotic environment that it is, this perfection can never be attained. So, how to advance?

By setting up the information streams that will allow the Governance decision-makers to know where they stand with respect to asset, project, and strategic management, and thereby know where their vulnerabilities are, and where their opportunities lie. The asset managers’ information feed already (and always) exists – all organizations need the general ledger, if, for no other purpose, than to pay taxes. But the idea that much, much more pertinent management information can be gleaned from the general ledger has been oversold, and to very bad ends. In addition to the general ledger, the Governance manager must have at his disposal an MIS that provides information on how the organization’s projects are performing, and the GL simply cannot provide this function. Nor can it provide the data feed on how the organization is doing in relation to its competitors. That information must be gleaned from sources that have nothing at all to do with the other feeds, but collected it must be. Once the information on how the organization is situated with respect to the three axes can be presented and evaluated can truly informed Governance decisions be made.

So, what does this model look like, and how do you set up the information streams that enable its functionality? Well, you can buy my book, or wait until Cameron asks us to re-visit the Governance theme.