Game Theory in Management

In the June 1952 issue of Collier’s magazine was a short story by famed Science Fiction author Ray Bradbury entitled “A Sound of Thunder.” Set in the year 2055, when time travel is available, big game hunter Eckels pays Time Safari to take him back more than 60 million years in the past in order to kill a Tyrannosaurus Rex. Travis, the guide from Time Safari, lays out some very strict rules for the hunters, including the need to stay on a levitating path that keeps the time-travelling hunters from interacting with the late Cretaceous environment. Travis points out that even the slightest change to the environment could have a cascading effect, impacting the hunting parties’ present. For those who haven’t read the story (or seen the film), SPOILER ALERT! that’s exactly what ends up happening after Eckels strays off of the path and accidentally steps on a single butterfly.

I find the use of the accidentally-killed butterfly to be fascinating, in that Edward Norton Lorenz would not describe the Chaos Theory building block, The Butterfly Effect, for another seventeen years.[i] The Butterfly Effect essentially posits that even very small perturbations within an interconnected network can have non-linear, even catastrophic impacts on the other parts of the network, often illustrated by the question “If a butterfly flaps its wings in Brazil, does it cause a hurricane in Texas?” Bradbury’s use of the butterfly image is a poignant way to offer up a treatment almost never seen in the multitude of movies and books that include time travel in their plots, namely that we really have no idea what would ensue if a particular historic variable were to be altered. Virtually all of the other works of fiction that deal with the question show the protagonists’ actions in their historic venues to have the effect of correcting a mistake in the time period from whence they came, with the theory of Unintended Consequences only making a rare appearance. It’s a deeply-held human conceit, I suppose, that we can possess a complete and comprehensive knowledge of the causal links that exists among events that occur in sequential order, and can correctly identify the impact of such decision points on their eventual outcomes.

Meanwhile, Back In The Project Management World…

After the newly-minted PMP® receives their certificate in the mail and has inserted it into a frame, the thought that there’s more to success in the PM world than what was taught in the certification-prep classes invariable enters their thinking. Probably the largest sector of this undiscovered country resides in the Organizational Behavior and Performance realm, which largely resembles Bizarro World from Superman. Are you encountering unreasonable resistance from Finance and Accounting to implementing even a basic Earned Value Management System? Bizarro World. Does every single Control Account Manager (CAM) insist that their work is strictly level of effort? Bizarro World.

But certainly one of the leading citizens of management’s version of Bizarro World is the representative of those holding to the idea that a complete and comprehensive knowledge of the causal links that exist among events that occur in sequential order can be correctly identified, along with the impact of such decision points on their eventual outcomes. Don’t misunderstand – I’m not knocking experience. Seasoned PMs are far more likely to make the correct decisions in an environment of incomplete information than rookies. No, it’s not the PMs I’m taking on here. It’s the Management Information Systems that are supposed to be informing those decisions. And the most common MIS used in the Project Management World that is predicated on the idea that the causal links that exist among events that occur in sequential order can be quantified and evaluated? That’s right, it’s the risk register, relentlessly pushed as essential by our old friends, the risk managers (no initial caps).

To prove my point, compare and contrast what happens during a typical risk event evaluation session with the examples available on-line, from major Universities, for calculating dependent risk. For the latter, examples usually include coin-flipping, dice-rolling, or other, limited-possible-outcome scenarios, which also have the advantage of having clearly mutually exclusive outcomes. One can’t pull anything other than a red or black playing card from a standard deck, for example.

However, in creating risk registers, subject matter experts are invited to offer up alternative events to those expected in the baselined management approach, and guess estimate both their odds of occurrence, and cost/schedule impact. These are then (usually) binned into categories pertaining to whether or not the events being considered are mutually exclusive, dependent in some way, or purely independent. But even this binning can’t be considered absolute, or even roughly reliable. Is there a risk event citing a 15% chance of increased vendor costs causing an 8% increase in budget? How can it be quantified, or even known, if this cost increase is passed along to any or all of the Project’s other suppliers? We’re not even remotely close to the type of closed systems where such calculations can be effective. Instead, the Project risk-evaluation environment far more closely resembles the levitating path for time-travelling Tyrannosaurus hunters, where the slightest change in parameters can have a cascading effect, rendering the very basis for modelling management decisions invalid.

And if accidentally stepping on a butterfly can utterly change the world, then just imagine what far more impactful things can do to just our projects.

[i] Wikipedia contributors. "Edward Norton Lorenz." Wikipedia, The Free Encyclopedia. Wikipedia, The Free Encyclopedia, 5 Jan. 2021. Web. 18 Jan. 2021.

In the Star Trek (the Original Series, or TOS) episode The Changeling, the starship Enterprise is struck by an explosive device fired from the deep-space probe Nomad in the opening scene. Nomad is apparently responsible for the deaths of over four billion inhabitants of the Malurian star system, and Science Officer Spock reports that the device that hit the Enterprise had the explosive equivalent of over 90 of their photon torpedoes. As a point of comparison, the energy yield from such a blast would be around 5,760 megatons, enough to power New York City (at its current rate of consumption) for approximately 1,677,768 years[i]. After several more hits, Captain James Kirk manages to communicate with Nomad, and discovers that it is actually a combination of an alien and Earth-originating probes. Its re-formed identity has led it to believe that it must eliminate anything it finds to be “imperfect,” primarily “biological infestations” (i.e., life) that it encounters, and it is headed to (where else?) Earth in order to carry out its twisted, horrific mission.

Of course, Captain Kirk manages to talk Nomad into destroying itself (fans of Star Trek are aware of the good captain’s talent for doing so), but before he accomplishes this we are treated to many lines of dialogue from the supposedly perfect thinking machine. One line that Nomad repeats several times is “Non sequitur. Your facts are uncoordinated.” A non-sequitur, of course, is a conclusion or statement that does not logically flow from the previous argument or statement.[ii] In other words, Nomad is alerting the speaker that they have offered up something irrelevant. The irony in this comes about when Captain Kirk informs Nomad that its belief that Kirk is actually Nomad’s creator, Jackson Roykirk, is itself in error and that, in order for Nomad to fulfill its programming, it must destroy itself. Heckuva price to pay for uttering something irrelevant, but then again Nomad was responsible for an awful lot of fictional alien deaths.

Meanwhile, Back In The Project Management Galaxy World…

Fortunately for we PM-types in the 21^st Century, offering up irrelevancies in the name of advancing management science in general, and PM science in particular, does not result in getting hit with a 5,760-megaton explosion, or even, it would seem, any condemnation at all. But management science is not advanced by chance interactions with extremely powerful alien probes combining elements of other management types’ techniques or practices. Indeed, I would go so far as to argue that the entirety of Project Management can be summarized as the quest to deliver accurate, timely, and relevant information to the people responsible for delivering a set piece of scope on-time, and on-budget. Whether or not a certain information stream is accurate can be objectively evaluated, and whether or not it’s timely is mostly objective, with a bit of subjectivity thrown in. But relevance? That’s a whole different probe animal. What’s relevant to one manager might be entirely superfluous to another, meaning that coming up with a single litmus test for relevancy is difficult in the extreme, if not impossible.

Muddying the waters here is the dual nature of Project Management Information Systems. Their primary purpose is to put into the hands of the aforementioned responsible decision-makers what they need to bring in their projects on-time, on-budget. Unfortunately, this perfect thinking machine deep space probe perceived itself to be incomplete, and had a chance encounter with the alien notion that PMISs exist in order to provide an audit trail for third parties to come in after the project is complete and criticize decisions with which the reviewers disagree. The terrifying conflation of these two management science notions has led to a state of affairs where four billion a whole lot of small projects that would have otherwise employed simple and effective PMISs, such as Earned Value-based ones, instead eschew them, dramatically lessening the odds of these projects reaching a successful close.

Fortunately for the PM Galaxy GTIM Nation, I’m around to point out the clear logical breakdown of the notion that PMISs exist primarily as a tool for ex post facto criticisms of the Project Team; however, I haven’t observed the spontaneous self-immolation of the opposing concepts, at least not yet.

Maybe I need to hit that idea with a blast equal to ninety photon torpedoes.

[i] Based on the yield of a single photon torpedo, from Memory Alpha (https://memory-alpha.fandom.com/wiki/Photon_torpedo) , and the rate of energy consumption of New York City, from The Leonard Steinerg Team, (https://theleonardsteinbergteam.com/2012/04/how-much-electricity-does-new-york-use/), and lots of unit conversions and number crunching.

[ii] https://www.google.com/search?client=firefox-b-1-d&q=non+sequitur+definition.

When I saw ProjectManagement.com’s theme of Resiliency for the month of January, I assumed it was meant primarily to encompass human resiliency, particularly since we appear to be required to continue to operate under less-than-optimal conditions in the PM world for the foreseeable future. But harken back to one of Hatfield’s Irrefutable Rules of Management, that the 80^th percentile best managers with access to only 20% of the information needed to obviate a given decision will be consistently out-performed by the 20^th percentile managers who have access to 80% of such information. This being the case, rather than point out and encourage the human factors that make up a resilient personality, I’d like to focus on something a bit more objective: which Project Management Information Systems are more resilient, or robust than others?

We can dispense with one of my favorite targets, risk management (no initial caps) right off the bat. Besides its inability to deliver information of any relevance, it is absolutely devoid of accuracy or precision, as demonstrated by the following thought exercise. Immediately prior to the flip of a fair coin, everybody knows that the odds of its landing one side facing up is fifty-fifty. Of course, after the flip, the coin has done one or the other, raising the question “Could a more precise prediction have been made?” Perhaps, if the one doing the predicting knows, with a high degree of precision, which side of the coin was facing up at the time of the flip, how much pressure went into the flip from both the thumb and arm motion, the precise points (I mean, like in microns) of the coin that were resting on the thumb and forefinger of the flipper, air density, temperature, weight and surface area of the coin, hardness or elasticity of the surface upon which the coin lands, exact height of the hand doing the flipping, etcetera, etcetera. Note that the absence of even one of these parameters immediately returns the predictor to the 50/50 call. Naturally, none of these data points are known at the time of the coin flip, or are even knowable to the level of precision needed for a greater than 80% accuracy, which is, ironically, the level of exactness called for in the infinitely more complex exercise of establishing a project’s contingency budget. Now, all that having been said, I suppose one could make the argument that risk management (no initial caps) is highly resilient, since its output is never held to any kind of an ex post facto evaluation, along the lines that, say, anyone could have predicted a 50% chance of a fair coin flip resulting in heads.

I’m not even going to touch Communications Management, since it doesn’t deal so much with information content as much as the manner and vehicles by which said information is conveyed to the people who need it to make informed decisions. I do, however, want to spend some pixel ink on our other friends, the accountants. I do not believe that double-entry bookkeeping can be said to be resilient. I think “brittle” is a better term, based on the myriad rules and guidance that goes in to Generally Accepted Accounting Principles. If any of those rules are violated, or an incorrect amount is entered into the General Ledger, then all subsequent output is thrown off by that amount. It’s why balancing the books is so critical to the month-end closing process. A balanced General Ledger passes my three tests for valid Management Information Systems (accurate, relevant, timely), but resilient? Not so much.

“So, Michael” I can hear GTIM Nation ask, “which MISs are resilient?” Well, the one that pops to my mind is Earned Value. EV has a remarkable self-healing capability. For example, let’s say that your project’s estimator provides you with a wildly optimistic cost baseline of $100,000 (USD) over ten months, and you begin work not knowing how comically low this figure is in reality. At the end of the first month, with the time-phased portion of the work worth $10,000, and the actual costs coming in at $10,000, everything seems to be going okay, right? Except your project controls analyst is showing only 5% complete, therefore $5,000 earned. The PM could be forgiven for believing the accountant’s take that everything’s on-track, even as the project controls person is predicting an Estimate at Completion of double the original cost baseline; however, after Month 2, with cumulative budget and actual costs figures at $20,000 each, but Earned Value still lagging at $10,000, the red flag is inescapable. The original estimate was disastrously low, the project is in real trouble, and none of the other information streams could have quantified, or even predicted the problem so early in the project’s life cycle. In an epistemological sense, the EV information stream was hit by the adverse event of a poor initial estimate, yet continued to provide an essential function without a hitch.

And that, GTIM Nation, is what I call a resilient information stream.

Since this will most likely be my last blog of 2020, and this past year has been, shall we say, less than optimal for goal, aspiration, and dream fulfillment, I think I’ll forgo any type of retrospective and focus instead on a certain quirk of PM cost/schedule performance systems used for projects funded on a Fiscal Year basis. Multi-year government projects are almost always in this category, as well as many funded by large organizations or institutions. A common problem with this type of work is that its Control Account or Work Package managers tend to want to avoid participating in Earned Value or Critical Path Methodology-based performance measurement systems, often times offering up the bromide that their work is “Level of Effort,” and therefore unsuitable for performance measurement. Yeah, dopey, but it happens all the time.

A common response from we PM-types when we hear this silliness is to point out the number of deliverables or measurable milestones associated with the work in question, thereby advancing the notion that the CA/WP could (and should) indicate its performance using one of the more discrete methods of assessing status. In my experience, the most common riposte to this challenge is to point to the fact that the CA/WP in question is estimated and re-estimated at the beginning of each Fiscal Year by projecting how much will be spent by the organizations that make up the members of the Project Team, which sort of points to the strange notion that the Projects exist to bring in funding to the organization, rather than the organization(s) existing in order to actually perform work. Again, dopey, but common. Due to this take on the relationship between Project and Asset Management, the assertion that the FY-funded work ought to be measured via a discrete method of collecting status almost never works.

So, what’s the PM needing a way of knowing project performance (other than sitting back and watching the actual costs go by) to do? One solution that I’ve discovered is remarkably simple – so simple, in fact, that it’s often derided by Earned Value or Critical Path purists as invalid. However, I can attest to its efficacy, returning usable cost/schedule performance information relatively quickly. The solution? Agree with them that it’s all Level of Effort, but with one small tweak: when the month-end status pull happens, ask the Control Account Managers if they are sure about that 8.333% figure.

See, 1/12^th is 8.333%, and if the project is being funded by Fiscal Year, then the LOE Control Accounts would nominally be 8.333% more complete each month. Explain to the CAMs that you know that they want to report progress based on LOE (where the Earned Value figure is set to match the cumulative budget amount), but they could really help out the PM if they could include some adjustments. Are you getting things done relatively quickly? Perhaps you could add a few points onto the status pull, and claim 8.5%, or even 9%. Conversely, if some snags have developed, maybe we could reduce the nominal claim down to 8.2%, or even just an additional 8% this month. I know it sounds miniscule, but I can show why it works using the Game Theorist’s favorite tool, the payoff grid:

As you can see, the only potentially bad outcome for the CAM is scenario #3, and yet even here the small tweak you are asking of the LOE Control Account Manager won’t strike them as raising unnecessarily provocative red flags. The tweaked amount is in tenths of a single percentage point, for pity’s sake. How massively alarming could that possibly be? And yet, tracking the cumulative Schedule Performance Index that is suddenly not anchored to 1.0000 yields its own highly relevant information stream, particularly when combined with a more accurate Cost Performance Index. If nothing else, the PM can narrow the internally-held Variance Analysis Thresholds, knowing that the proffered variances are most likely to be understated. The cost/schedule performance system has new life, and the PMs in charge of Fiscal Year-funded work finally have the coveted disaster early-warning system, the lack of which produces a seemingly never-ending stretch of sleepless nights.

And all because someone asked the question, “Are you sure about that 8.333% complete figure?”

There’s an axiom among experienced writers that asserts you’re not really a pro at this until you make the transition away from writing something that you want get off of your chest, and towards producing something that your readers would actually want to read. I like to keep this axiom in mind when I’m attending conferences on the management sciences. It serves as a valuable litmus test when I’m reviewing the syllabus, to determine which of the sessions is worth my time, and which should be avoided. In those instances where all of the slots in a given time frame are clearly presenters attempting to show off, boast, or re-address intellectual ground that’s been covered so many times it’s taking on the epistemologically equivalent appearance of centuries-old roads used by the Roman Legions, I use that time to check out the exhibit hall, or to get in line early for lunch.

I kind of get a kick out of speculating what would happen if all of the attendees at these conferences were to employ a similar test when choosing which sessions to attend. The boasters and re-treaders would speak to empty chairs, while the ones who have put real effort into assessing a rational approach to problems or issues encountered by many would see their rooms filled to overflowing, thereby sending (one would hope) a clear message to the seminars’ sponsors in general, and their paper-selecting committees in particular: stop with the common knowledge re-iterations, cease the a priori­-predicated yawn-fests, and let your valuable time and space allocations to those who have positioned themselves to provide actual value to potential attendees.

But in order for this message to arrive in the in-boxes of seminar sponsors, we as a PM community need to do something about the supply and demand curves involved. Since it’s pretty clear that there will always be a large supply of potential paper presenters who are there to boast about some wonderful project involving them, and point to some unique aspect of the PM codex as a distinguishing factor that all other analogous projects should adopt, my mission is probably best accomplished by flattening the demand curve. I believe this could be accomplished by employing a few tactics, such as:

• The aforementioned pre-attendance screening, performed by seminar attendees. Maybe we could get a rating system – we’ll call it the GTIM Nation Relevance Score – that would show a numeric value of the subject presentation’s basis in management science techniques.
  • 10 – 7 would be heavily based on either existing, hard-evidence-based scholarship, or original research that includes verifiably objective data, with a minimum of subjectivity, or correlation masquerading as causation.
  • 6 – 4 means that we’re going back over already-established techniques and protocols, with a heavy dose of eat-your-peas-style hectoring about why it’s important to perform them.
  • 3 – 2 score indicates subjective data is being used to draw conclusions (polls and surveys are a dead giveaway), or else objective data is being process through an invalid logical method to assert a pre-determined conclusion. Unless I’m mistaken, this would be the automatic scoring range of every risk management (no initial caps) session I’ve ever attended.
  • 1 – 0 is for straight-up silliness, like the Communications experts who want Project Teams to share all project information with anyone who can be considered to be a “stakeholder.”
• I suppose it’s possible that the review committees will have some sort of epiphany (perhaps their members will be sent links to this blog), and begin evaluating paper/presentation proposals on the basis of the actual validity of the papers’ premise, and supporting evidence as the sole basis for awarding slots.
• I believe that there’s a role in all of this that the people actually sponsoring the events can play, but it’s not a popular idea. I had attended a PM-related seminar, where I gave a keynote address (this was back when I was doing the PMNetwork column gig), and was invited to a post-conference meeting with the sponsors and other presenters. The purpose of the meeting was to generate some sort of lessons-learned memorandum, to help improve the same get-together the next year. When they opened the floor to suggestions, I raised my hand, was recognized, and blurted out “If you really want to improve this thing, remove all attendance fees from paper presenters.” At the time those who had submitted presentation proposals that were approved would have their attendance fees reduced, but not eliminated. I went on to point out that ideas were the coin of the realm in these venues, and, in order to attract better ones, all barriers needed to be eliminated. Like I said, this was not a popular idea.

As much as I would love to believe that a simple ProjectManagement.com blogger railing against the intellectually vacuous underpinnings of what all-too-often passes for legitimate Project Management science could tip the common acceptance scales against them, that’s not gonna happen. On the other hand, if we as PM technique consumers simply stop signing up for what these people are selling, now that would turn the tide. The only question remaining is: have we had enough?