Game Theory in Management

Here in March 2021, the one-year anniversary of the imposition of pandemic-avoiding regulations, it’s somewhat natural to reflect back on what we’ve experienced and what we’ve learned during this time. In any discussion of the management sciences, though, the question of what we’ve learned from any given history must come with the acknowledgement that we’re never dealing with a purely experimental setting. There are simply too many parameters involved whenever we attempt to draw a usable inference of the causal elements that went into any particular management strategy.

The Covid-19 pandemic, however, represented a large-scale, broad-based condition on the macro economy. In other words, as negative as it was and is, it had some level of impact on virtually everybody, and on the way everybody conducted business. Along the way the past year has shown some rather obvious problems with several previously-widely accepted theories, such as the ones I addressed in last week’s blog, but it has also shown a light on some management strategies that were perhaps given short-shrift prior to 2020. For example, once the restrictions on business began to loosen, what were those business’ main concern? Were they associated with “maximizing shareholder wealth,” or were they oriented towards getting their customers back? And, before anybody tries to assert that the former encompasses the latter, let me be more specific: in the various stages of businesses being allowed to re-open, did anybody really care about the return-on-investment figures for purchasing or leasing the new copier?

While the Asset Managers’ (read: accountants’) main concern is, well, the performance of the organizations’ assets, the ability to deliver for customers belongs to the PM world. None of the information streams that inform management of the organization’s progress towards accomplishing the customers’ expectations for achieving scope, cost, or schedule performance can be relayed via the general ledger (and don’t get me started on the output from risk management [no initial caps] “analysis”). That’s simply a fact. The only reason that the belief that a project can be successfully managed without those pesky PM artifacts like the Work Breakdown Structure or Earned Value Management System advances is that projects that never had anything go wrong with them on occasion relied exclusively on the general ledger to inform their decisions, so it follows (erroneously) that all successful projects don’t need those things.

But what happens when something does go wrong with the project? As I pointed out in the previous paragraph, what’s needed is a Management Information System (MIS) that can answer a whole bunch of questions, and the general ledger isn’t that system. What questions? Well, ones like:

• Is the project really performing poorly (I mean, past my vague feeling)?
• If so, is its performance lacking in cost, schedule, or both?
• How big is the problem?
• Where, specifically, is the problem? What part of the project is experiencing it?
• What are the implications long-term? In other words, what is going to happen if the problem isn’t addressed, or isn’t addressed fully or appropriately?
• What’s causing this problem?

While the general ledger can only marginally or tangentially answer any of these questions, an Earned Value Management System, based on a valid WBS, can answer all of them, with excellent precision. Throw in a Critical Path Methodology capability, and the resulting information stream becomes all the more powerful. Allow me to point out again that such information streams only become vital when something goes wrong. A perfectly healthy patient does not need an X-ray, and could probably spend his whole life not knowing which nearby medical facilities have a machine capable of performing such imaging without issue. That all changes real fast if such a person breaks an arm.

This all has the effect of placing much of the motivation for initiating and maintaining Project Management Information Systems in an organization to those analogous to purchasing insurance. If someone is supremely confident that they will never get sick or break an arm, they may tell themselves that it’s a waste of money to buy insurance. Similarly, if a (so-called) PM is confident that they “manage” so well that they don’t really need a performance measurement system to inform their decisions, they may forgo the cost of setting up such a system. All of which brings us back to the title of this blog, the famous PM saying that the reason project teams detest performance measurement systems is that they vividly show their lack of performance.

Even if the downturn in performance was caused by a global pandemic.

Have you ever been in a situation where your employer has requested/directed/demanded that you put in extra hours, but upon putting in the extra time, no positive result was observed? The notion that organizations perform best when their Return on Investment figures go up is a core concept in business models everywhere, so I suppose it falls to me to show how it’s highly suspect at best, counterproductive often, and has no place in the realm of reflexive management tactics.

The error falls into two categories, one from our friends, the Asset Managers, and the other from a popular but, in my opinion, intellectually vacuous novel based on a derivative of a well-known and established PM technique. First, the ability of Critical Path Methodology (CPM) software to correctly identify which activities are directly responsible for potential late scope completion, as well as the organizations performing those activities, has to be one of the most powerful pieces of management information that PM-centric systems can generate. Management time and energy is finite. The ability to know which parts of the project and team that are doing just fine without managerial input and those that need direct attention is huge, and can go a long way towards optimizing PM’s time. However, this information is only available from CPM-capable systems, and not from the Asset Managers’ main tool, the general ledger. It follows, then, that Asset Management-based solutions do not work on Project Management-oriented problems, like overcoming a potential late project delivery date, or milestone. The unfortunate put-upon staff from the question in the first paragraph getting called in to work may have done absolutely nothing to help the project get back on-time, but somehow the notion that these resources were “working” without incurring any additional marginal costs (assuming they were all on salary) makes this demand seem somehow legitimate. In reality it takes a wrecking ball to morale; but, since morale can’t be quantified, this error is rarely criticized or condemned.

The second error category that pertains to this blunder has to do with a work of fiction from 1997, entitled Critical Chain, by Eliyahu Goldratt. In the novel, the protagonist manages to make progress towards schedule goals by using a technique that seasoned schedulers would instantly recognize as “crashing the schedule.” Crashing the schedule involves assigning more resources to critical activities – particularly ones that are in or may become trouble – in a bid to accomplish the scope more quickly.

There are several attendant problems with crashing the schedule beyond increased costs or the possibility that the increased resource density could actually decrease productivity, but the main issue with the tactic has to do with an assumed commonality of expertise. If, say, your electricians look like they won’t finish on-time, it really doesn’t do any good to call in the cement pourers. In the example above, the people being called-in to work needless overtime didn’t necessarily improve schedule performance. These problems with the push-staff-harder approach, however, did not stand in the way of the rebranding of crashing the schedule as “critical chain” management becoming something of a sensation in PM circles. There’s even a Goldratt Institute, which probably owes no small part of its existence to the success of the 1997 book. But that’s kind of the thing with PM theories that may or may not work so well in the real world: they can always, always be made to work in the world of fiction. It’s highly reminiscent of B. F. Skinner’s 1948 novel Walden Two, where his then-nascent theories of behaviorism are used to govern a small but extraordinarily well-ordered society, where virtually the entire population is happy and performing at near-peak potential fulfillment levels. Behaviorism would go on to become a major school of psychological thought before finally receding in popularity during the “cognitive revolution.”[i] While asserting that, as a psychological trope, it was blown to smithereens may be a bit excessive, I would go so far as to say that any near-science hypothesis – like those belonging to psychology or the management sciences – that achieves widespread acceptance not through experimentation and the publishing of results, but via fiction, is fully deserving of higher scrutiny, if not reflexive abandonment.

As for the notion that automatically pushing staff to work harder whenever a schedule setback is encountered, yeah, that’s one management trope that should be blown to smithereens.

[i] Wikipedia contributors. (2021, March 6). Cognitive revolution. In Wikipedia, The Free Encyclopedia. Retrieved 01:33, March 9, 2021, from https://en.wikipedia.org/w/index.php?title=Cognitive_revolution&oldid=1010625203

Depending on one’s particular flavor of PM, much of the data needed to make informed decisions can be gleaned through means other than donning a hard hat and walking the site, or showing up to the Weekly Scrum in-person (masked and distanced, of course). With the adjustments made to virtually all of our business models to accommodate the pandemic, several hypotheses of management science have rightfully made progress towards acceptance as theories, while some widely-accepted theories that have served as the basis for many analysis techniques have been shown to be suspect at best, and openly fraudulent at worst.

In the former category, I offer up as Exhibit A that it’s the Earned Value Management Systems that represent the optimal tool for assessing the impact of a negative macro-economic event, and not the general ledger. The reason for this is simple: an EVMS can readily document the delta between an organization’s pre-COVID performance and what happened afterwards. Of course, a little bit of binning is in order. Once a usable baseline of Project performance can be established (say, its cost/schedule behavior for the three months leading up to the shutdown), then a comparison of its execution afterwards with such a baseline will yield the desired impact information, so:

Note that in only half of the scenarios above should a documented change in cost/schedule performance be attributed to the macro-economic event.

Now, if Earned Value has been shown to be the most appropriate tool for capturing the cost and schedule impact of a macro-economic event, what did it displace? Traditionally, it’s our friends the accountants who have been the go-to team for answers on any and all questions where the solution involved a dollar sign. Unfortunately, the only way to attempt to glean pandemic-related impacts from the General Ledger would be to examine all expenditures, and try to estimate some causality-related connection as a precise number. A few honest questions will show how utterly impossible such estimations can be:

• Should any increase in costs of a budgeted line item be attributed to the shutdown?
• If yes, then by what amount – the amount of the delta, or should some factor be included that takes into account other influences, of which there are multitudes? Also, should any decrease be subtracted from the overall impact factors?
• If no, then by what other GL-owning data could impact numbers be derived?
• What happens if costs went up, but so did actual performance? Oh, wait, the GL has no way of capturing that, so never mind.

I could go on (and often do), but GTIM Nation sees my point.

As for which commonly-invoked theories have been hit with serious challenges to their efficacy, I think the most blatant is the use of Return on Investment as the ultimate arbiter of “good” and “bad” in the management decision-making universe. Inanimate assets don’t perform. People, usually collected into – what’s the term I’m looking for, oh, yeah, Project Teams – perform, usually taking advantage of the other assets at their disposal. The printer doesn’t perform, the person creating the document does. However, in supposedly first-rate business schools around the globe, the analysis method for determining which projects to pursue within a portfolio is its anticipated ROI. Even a cursory review of the formula for generating the ROI shows it to be so chocked-full of subjective (or even crazy speculative) data that it makes our friends, the risk managers (no initial caps) seem positively insightful.  So, when the shutdowns descended upon us, a large segment of the parameters used to assess ROI, such as the need for employees to use office space, were shown to be, rather abruptly, utterly irrelevant.

Savvy Program Managers will have quickly recognized the sea-change in the menu of acceptable management science theories in these times of duress, both those added, and those that should have made an exit long ago.

Others? Not so much.

[i] Cost Performance Index (CPI) is the Earned Value divided by Actual Costs, and the Schedule Performance Index (SPI) is the Earned Value divided by the cumulative time-phased budget.

As GTIM Nation is aware, my litmus test for Management Information Systems’ validity (including the PM variety) is that they have three characteristics:

1. They must be timely. Regardless of the exact information stream, it’s going to have a shelf life, often measurable in days if not hours.
2. It must be accurate. Wrong information stemming from poor data or wrong-headed processing techniques is worse than a simple waste of time and energy – it’s almost always misleading.
3. It must be relevant. Since information takes time and energy to produce, irrelevant data is automatically a waste of time, though it can possibly be of value, the same way a broken clock is correct twice a day.

So, when the discussion turns to the effect of digital transformation on PM (ProjectManagement.com’s theme for February), I take such transformations to mean their effects on the information streams that guide optimal managerial decision-making. Here another one of Hatfield’s Indisputable Laws of Project Management comes into play, specifically that the 80^th percentile best managers with access to only 20% of the information needed to obviate a given decision will be consistently outperformed by the 20^th percentile worst managers who have access to 80% of the information so needed (my little PM derivative of the Pareto Principle). Taken together, the PMIS Litmus Test and Hatfield’s PM Variation of the Pareto Principle form the basis of an assessment of the beneficiaries of digital transformation in Project Management.

Let’s start with evaluating advances in data processing with respect to PMIS’s accuracy. Does it help, or hinder? I would assert neither, really, save an ability to more promptly detect when serious inaccuracies creep into the data. The creation of the original scope baseline remains largely unaffected by digital advances, save the leap from using typewriters to word processors in the generation of the documents that hold the scope. Yes, estimators haven’t had to use ledger paper for decades, but their basic arithmetic remains largely unchanged. Also, when the Project Controls Analysts collect status, actual start and finish dates are what they are, and estimations of percent complete are similarly unchanged since the early days of Earned Value and Critical Path Methodologies.

Has digital transformation had an influence on the relevance of Project Management Information Systems? Yes, but not necessarily a good one. While the EVM and CPM-based reports retain their relevance from the days that computers received their instructions from punched cards (and if you GTIM Nation whippersnappers out there have no idea what I’m talking about, I don’t want to hear from you), other so-called information streams have been making inroads into the commonly-accepted PM codex, though they should not have. Running three or four highly speculative alternatives to the planned course of a given project through a Monte Carlo analysis (made much easier through the digital transformation of PM) does not render its output any more relevant. If the output curve is pre-selected, as it is in almost all risk management (no initial caps) software systems, then the number of “random” data points added is truly irrelevant – the outcome has already been decided. The added data points are just there for show.

Which brings us to timeliness, which is where the greatest amount of added value to PM can be seen. In the distinction between feedback and feed-forward Management Information Systems, each type has its advantages and vulnerabilities, specifically:

• Feedback systems use only verifiable, objective data, meaning that they are more accurate. However, since data takes time to collect and process into usable information, Feedback systems are vulnerable to delays in delivering their information (see bullet #1 in the first paragraph).
• Feed-forward systems, conversely, are predicated on highly subjective data, like the perception of historical trends continuing into the future, or even more suspect methods such as risk analysis (no initial caps). This type of system is more timely, but significantly less accurate.

With the digital transformation of PM delivering processed information far faster than was possible previously, the hands-down winner of the most-benefitted award has to be the Feedback systems. Helping Feedback systems overcome their timeliness vulnerability was always something that could be accomplished by advances in more advanced data processing, but nothing in such advances could ever lend accuracy (or, if we’re being honest, even relevancy) to the Feed-forward systems. Which brings us back to the question posed in the title, Who benefits from the digital transformation of PM? Ironically, it’s those of us who have been stressing the primacy of Earned Value and Critical Path Methodologies-based information streams.

In other words, it’s largely the members of the PM community who know what a computer that uses punched cards looks like.

I recall seeing one definition or aspect of human intelligence described as the ability to identify analogous situations to those presently encountered, and being able to appropriately adjust known strategies from those analogous situations to derive an effective response or strategy to the new, novel one. Assuming this assertion is at least partially accurate (and I believe it is), it leads to quite the conundrum when it comes to artificial intelligence, or even the impacts of digital transformations (ProjectManagement.com’s theme for February). Recall that the root of all digital transformation is the bit. Computers exclusively use the bit to perform all calculations and evaluations, with no exceptions. A bit is either a zero or a one, an on or an off. Bits are collected into sets of eight known as bytes. In Random Access Memory alone, the computer I’m using to write this blog has 32 gigabytes, or 257,698,037,760 bits, and those data points do not address calculations per second, disk capacity, or any of the other parameters used in evaluating digital computing performance. It can perform virtual reality, web surfing, media playing, word processing (obviously), and hundreds of other pretty amazing things. Still, it all comes down to the bit – on or off, one or zero, is or is not.

Which brings us back to our conundrum. Project Managers may seem to have a rather simple set of tasks to execute, and the proper way of executing them has generated more guidance than could probably fit on my hard drive (2.72 TB, or 2,989,847,736,320 bytes). I myself have written about some of the more robotic aspects of filling in the Corrective Actions section of a typical Variance Analysis Report, to wit:

But if the events and circumstances surrounding the decisions that we PMs make on a day-to-day basic could be reduced to an all-inclusive set of objectively measurable parameters, then computers with the ability to evaluate Earned Value and Critical Path data and connect that information stream to a (much larger) response codex similar to the table above would have replaced us long ago. That aint gonna happen, at least not anytime soon, and the reason should be obvious: there’s far too much nuance attached to the PM decision-making process than could ever be reduced to a formulaic when-you-see-this-do-that response system, and any attempt to create such a system (cough, risk management, cough) can only proceed from a rather dubious reductionist starting point.

This aspect of PM serves as part of the answer to the question I posed in last week’s blog, on why digital transformation, while having a profound impact on so many other areas of human endeavor, haven’t had a similar impact on PM. Projects are, by definition, unique. Sure, some aspects of PM can be executed via template, like the list of rules for creating a Work Breakdown Structure[i]; but those aspects, ignored or abused as they often are, are in the minority of the matters and issues requiring insightful decision-making from the PM. As cringe-worthy as I find the assertion that PM is as much art as science, I cannot flatly refute it, as much as I may wish to.

All of which points to our epistemological dichotomy, between those parts of PM that can be addressed as Boolean choices – digital, if you will – as compared to that set of decisions within the purview of PM that are so nuanced that not only can they not be adequately addressed via some codex of hard-and-fast rules, they probably can’t even be solidly justified by their associated audit trails. Red-light cameras at intersections may be able to take one particular enforcement of a traffic ordinance off of a police officer’s responsibilities, but would you want such a device calling an automatic infraction for an improper lane change, much less a far more complex violation?

So, yeah, a digital transformation doesn’t necessarily mean that the thing being transformed is getting smarter. It might not even be intelligent.

[i] Basic rules for setting up a WBS: (1) each WBS element must have a set piece of scope, (2) discernible beginning and ending dates, (3) a specific set of resources assigned to it, (4) one person (or organizational entity) is responsible for it, and (5) no “child” can have more than one “parent.” Without these conditions fulfilled, odds are you’ve placed an Organizational or Functional Breakdown Structure element into your WBS.