Game Theory in Management

When an author attempts to advance a theory under the auspices of Management Science using obvious sophistry, I immediately know two things:

• The theory has not been proven, or even logically advanced past the hypothetical phase, and
• Any practitioner of the legitimate hard sciences is fully justified in rolling their eyes clear to the back of their heads whenever they hear the term “Management Science.”

Of course, many instances of sophistry are difficult to detect, at least right away. Not so the case with the straw man argument, no siree. This is where the theoretical position opposite the assertion being advanced is set up, but not accurately, then criticized to oblivion. And this is exactly the tack being taken with much of the advancement of earned schedule (like “risk management,” I refuse to use initial caps for this term).

What is “earned schedule?” Well, from the self-proclaimed “official site for Earned Schedule,” we see this:

Earned Value Management (EVM) is a wonderful management system, integrating in a very intriguing way, cost …schedule …and technical performance. It is a system, however, that causes difficulty to those just being introduced to its concepts.  EVM measures schedule performance not in units of time, but rather in cost, i.e. dollars.  After overcoming this mental obstacle, we later discover another quirk of EVM: at the completion of a project which is behind schedule, Schedule Variance (SV) is equal to zero, and the Schedule Performance Index (SPI) equals unity.  We know the project completed late, yet the indicator values say the project has …perfect schedule performance!![i]

In just these five sentences we can begin to see the vacuousness of its underpinnings. Quick question: why does the author feel the need to refer to EVM as “wonderful” and “intriguing?” Also in the first sentence, the author doesn’t understand how to use ellipses, twice interjecting them for commas, and don’t get me started on multiple exclamation points. Does that last objection sound like nit-picking? Maybe so. I just have a hard time taking advice on how to better manage hundreds of thousands of dollars’ worth of projects (if not millions) from a person who hasn’t mastered eighth-grade English.

The third sentence points out that Earned Value Management (EVM) measures schedule performance in units of cost. Umm, yeah, that’s what EVM does. If you want to measure schedule performance in terms of time, which earned schedule claims to do, then the traditionalists would turn to – Critical Path! That’s what CPM does. In another document, generated by a certain guidance-document-generating organization that I refuse to name, the example given for earned schedule has to do with reducing plans to meet friends for dinner to PM terms. The story problem stipulates that, at your current rate of performance, you will be late to a dinner with friends, and goes on to mock the idea that you would contact them to announce that you will be X “dollars late.”

But this is a straw man argument, and, therefore, invalid. When an EVM system returns that one will be $X late, it doesn’t mean that time equals money (though some would argue the contrary). It simply means that, if you want to arrive on-time to your dinner date, you should be prepared to spend $X over and above what you had originally budgeted to make that happen. Variances at Completion expressed in units of time are not properly derived from an EVMS – again, they come from Critical Path Methodology systems. It is, in fact, their raison d’etre. And yet the earned schedule crowd seems to assert that this new technique has bridged some previously-unsolvable management information gap. It’s simply not so.

Another straw man argument from the excerpt above challenging EVM and its schedule performance metric has to do with the fact that the Schedule Performance Index (SPI, or the cumulative Earned Value amount divided by the time-phased budget), in a “quirk,” moves toward 1.00 as the project comes to completion, regardless of whether or not it is finishing on-time. Again, this is an irrelevant observation. The SPI simply measures progress against the baseline; of course it’s going to close in on 1.00 as the project nears completion. There’s no “quirk” about it. You want a performance measure that compares a projected finish date to the original baseline date? That comes from the CPM system, and for experts to blithely fail to take this into account is sophistry.

I understand that the concept of earned schedule has received a lot of attention and accolades in the PM world. So has risk management. So has communications management. I also understand that it only took a random Cairn Terrier to pull back the curtain on the Wizard of Oz.

[i] Retrieved from https://earnedschedule.com/ on April 25, 2021, 18:45 MDT.

In the Steve Martin movie The Man With Two Brains, Martin’s character, Dr. Hfuhruhurr (you’ll need to see the movie to hear how it’s pronounced) is pulled over by the Austrian police while driving at excessive speeds. They perform a field sobriety test that includes the following steps:

• Produce license,
• Get out of the car,
• Stretch arms out to the sides, and then touch nose,
• Walk the white line (lane demarcation),
• Come back, but while walking on hands,
• One hand,
• Perform a flip,
• And simultaneously juggle three oranges, tap dance, and sing the “Katalina Matalina” song,

…all of which Dr. Hfuhruhurr performs successfully, while commenting “&^%$* your drunk tests are hard!”

Prior to seeing this movie I had never heard of the Katalina Matalina song, but I understand it’s fairly familiar to school-age children. The chorus goes like this:

Katalina Matalina

Upsidina walkadina

Hoca poca loca

Was her name.

The verses are hardly better. I would go on, but I’ve probably already sunk the Flesch-Kincaid Grade Level Readability Calculator score for this blog to levels so low that the ProjectManagement.com editorial staff may automatically reject it.

Meanwhile, Back In The Project Management World…

As our Earned Value and Critical Path Methodologies (EVM/.CPM) cost/schedule control systems go hurtling down the Project Management Information highway, they will sometimes attract the attention of PM constables who will pull them over and politely but authoritatively ask them to perform a few simple tests to determine their validity. It’s easy to see how a lot of these tests go directly to system efficacy, such as:

• The presence of an activity percent complete without an actual start date,
• An actual finish date without 100% complete,
• Activities that have Finish-to-Start logic ties to each other (circular logic),
• Negative float (indicating at least one activity on the critical path has an assigned constraint),

…among others. And, while the PM constables are usually very polite, the clear implication is that, should the Project Management Information System being evaluated happen to fail these tests, the take-away would be that the Project Team is either incompetent, deceitful, or both.

Aiding these PM constables in their duties are software tools that can scan large CPM networks or EV systems. This is all well and good, but I have to ask: what happens when a one-size-fits-all software intended to check system integrity is run against the Earned Value or Critical Path Methodologies-based systems associated with a-typical projects?

Here’s the situation using the Game Theorists’ favorite tool, the Payoff Grid:

In Scenarios A and D, the system integrity-checking software has performed as intended, and needs no further evaluation. However, if there are genuine problems with the PMIS being evaluated, and the software doesn’t pick up on it (Scenario B), then it looks really bad for that package, particularly if the subject project ends up overrunning or coming in rather late, with no early warning from the PMIS. In those instances where the PMIS is really okay, but the software came back with a list of errors, the natural inclination is for the Project Team is to chase those to ground. Prior to this forensic analysis and pursuit of the remedies, one question should be asked: are all of the checking software’s tests relevant?

Consider, for example, the old saw about how, if the number of activities (or percent of their budgets) using the Level-of-Effort method to claim their Earned Value amounts is over anywhere from 5% to 15%, this is indicative of error. I understand the value of using the more discreet methods of claiming EV, such as direct units or weighted milestones, when plausible. But for those projects that are more service-oriented than others, LOE will almost certainly be the most appropriate EV method for a plurality, if not a majority of its activities. Evidence for this assertion lies with the fact that, ironically, the Project Management task in almost all projects is invariably tracked using LOE, meaning that it’s entirely possible that the PM consulting firm performing a baseline integrity review using one of these software packages wouldn’t get a passing grade for their own PMIS.

Don’t misunderstand – I’m all for PMIS integrity, and for the software tools available that can help attain it. I’m also in favor of the detection of drunk drivers in Austria. I just think that having to walk on my hands on a straight line to establish sobriety is a bit excessive.

Besides, I don’t even know all the words to the Katalina Matalina song.

Regular members of GTIM Nation know of my disdain for risk management (no initial caps) as currently practiced; however, there is no truth to the rumor that I stated that it is a total and complete waste of time, it only serves to muddy the Management Information System (MIS) waters, or that I have compared its practitioners to members of the genus Mustelidae. I have, though, regularly stated that it fails two of my three criteria for valid management information systems, that they be:

• Accurate,
• Timely, and, perhaps most important of all,
• Relevant.

Being the traditional kind of PM that I am, I think it’s clear that the standard methodologies of Earned Value and Critical Path represent valid systems, while risk management[i] doesn’t. But curmudgeonly blogger talk is cheap: what can we see in legitimate management science space that would convincingly point to the conclusion that risk management[ii] methods are objectively inferior to, say, Earned Value?

To set up this test, let’s first find a common output from each system. Risk management (I only used an initial cap on the word “risk” because it started the sentence) cannot tell you:

• Cost variance,
• Schedule variance,
• Which Control Accounts or Work Packages are doing okay in cost/schedule performance space,
• …and which are in trouble
• …or by how much,

…all of which, the alert reader will realize, are highly relevant pieces of information. Conversely, Earned Value Management Systems cannot tell you:

• Odds of speculated alternatives to the cost baseline actually occurring,
• …or their estimated impact,
• The appropriate amount to set aside for a contingency budget,
• Or the “confidence interval” that the original cost baseline will not be exceeded,

…all of which, the alert reader will realize, are fairly irrelevant, with the possible exception of the third bullet (but even that is highly dependent upon the ability to accurately capture its underlying assumptions’ data, which is itself suspect). So, what relevant piece of information do both types of systems assert an ability to generate?

“I’ll take ‘Variances at Completion’ for $1000, Alex.”

 To be sure, this apparently crucial piece of information doesn’t come by the risk managers[iii] easily. In order to provide an accurate prediction of how much a given project will cost when it is complete, how long it will take, and compare those figures to the original baseline estimates, the following processes and consideration must come into play:

• Assuming that the project’s original scope baseline was used to create its cost and schedule baselines, and that these latter two have a nominal “confidence interval” of 80%, it then follows that the amount of contingency derived from the risk baseline represents a not-to-exceed limit.
• Further assuming that the risk analysis was performed at an appropriate level of the Work Breakdown Structure (WBS) – one would hope for the Work Package level, but Control Account-level is perhaps usable – the odds of each of the alternative scenarios to the one described in the WP are multiplied by the estimated impact (both cost and schedule), and summed at that same WBS level.
• If any of the alternate scenarios occur, the amount of contingency assigned to that alternative is added to the nominal Budget at Completion to derive a new Estimate at Completion (EAC).
• On the other hand, should a task receiving this type of analysis actually finish without a risk event occurring, the amount of contingency associated with that task is “released,” and the EAC remains unchanged.
• If an event that impacts the project’s costs, unexpected by both the PM and the risk managers, occurs, it is classified as an “unknown unknown,”[iv] and its quickly-estimated impacts are added to the previous EAC.

Note that these steps are not scalable. The alternative scenarios either happen, or they do not. At that point, the entirety of their estimated impact is the “right” number, or it is not. The act of multiplying the impact amount by the speculated estimated odds of occurrence only provides useful information in the aggregate. Note also that most of these steps require the time and attention of some fairly proficient analysts.

So, how would an Earned Value system provide this same information?

• Divide the estimated percent complete figure into the cumulative actual costs. The same thing works for duration – divide the percent complete figure into the number of days since the project’s start date.

Note that these steps are perfectly scalable. They will return a figure accurate to within ten points at whatever level of the WBS is being assessed. Note also that it requires no special expertise (to perform – setting up the original baselines do require some level of competence, but would need to be done for the risk managers anyway). A grade schooler could do it.

So, I put it to my readers, both Members of GTIM Nation and occasional: which information stream do you think is superior?

[i] No initial caps.

[ii] Ibid.

[iii] Ibid.

[iv] …which has to be one of the goofiest terms in all of management science.

In the television series Star Trek, The Next Generation (TNG), one of the main characters throughout its run was Lt. Commander Deanna Troi, played by Marina Sirtus. She was half-human, half-Betazoid, a species with telepathic abilities. As “ship’s counselor,” she occupied a chair on the bridge of the Enterprise, right next to Captain Picard, opposite First Officer Riker. With such prominent placement on the bridge of The Federation’s flagship, one would think she had an awful lot to contribute. And one would be mistaken.

Just as the Star Trek, The Original Series’ (TOS) character Commander Spock had the familiar line “most illogical,” or “fascinating,” Lt. Commander Troi is probably best known for her phrase “I sense he’s hiding something, Captain,” usually when such an observation could have easily been made by any of the non-telepathic members of the bridge crew, and almost always when the subject of her remark was obviously obfuscating. To be fair, there were a couple (at least) of episodes where her telepathic abilities provided an insight as to the true motives of that particular episode’s antagonist, but for the most part I found her intuitions to be clearly redundant.

Meanwhile, Back In The Project Management World…

While not telepathic in nature, there can be no doubt that we PM-types have access to information and insights completely foreign to those outside our clan. For example, consider a grand assembly of the principals of a given organization, meeting to ascertain, well, the usual question on everybody’s mind: how are we doing? Of course, each specialty will interpret the question, technique for deriving an answer, and appropriate response very differently, so:

• Our friends, the Accountants, will want to answer the question by providing information from the general ledger, specifically the profit-and-loss statement. If they do foray into project performance space, it will be to project an Estimate at Completion predicated entirely on the rate of spending, by project.
• Our other friends, the risk managers (no initial caps) will turn to the risk register, add up the previous risk events that they had guessed foreseen that actually came about, summarize their collective costs, and compare that to the total time-phased contingency budget across the projects in the organization’s portfolio to see if risk-event-related costs are out-pacing their budgets. In the alternative, they will compute the approval rate of all of the Baseline Change Proposals submitted due to the occurrence of any risk event, on the risk register or not, to gauge the individual projects’ ability to compensate for said events and that are likely to come in on-time, on-budget.

Just kidding. Expect them to assert a judgement on organizational health based entirely on the projects’ willingness to pay them to set up a risk register.

• Our still other friends (GTIM Nation makes a lot of friends, don’t we?), the Communications Managers, will gauge organizational health by performing an assessment of the ability of all “stakeholders” to weigh in on key decisions in project space. They’re always searching for that one overlooked stakeholder who has the answer, but hasn’t been asked nicely to interact with the organization’s principals.
• Expect the Quality Managers (we can’t be their friends, due to their high standards) to proclaim acceptable (or not) organizational performance based on customer perceptions of the organization’s output, regardless of whether or not this output is consistent with the original scope baseline.
• The rogue estimators (make no mistake: if the estimators don’t belong to the PMO, they’ve gone rogue) will appear to come closer to the relevant information for this meeting by re-estimating the work remaining in the projects in the portfolio, adding on the cumulative actual costs, and proclaiming reliable Estimates at Completion. (Narrator: the numbers they generate are not, in reality, reliable.)
• Finally, we have the Project Management-types, who can predict with unmatched accuracy which projects in the portfolio will overrun, underrun, come in late or early, or any combination thereof, and by how much. Even the most basic of Earned Value Management Systems (EVMSs) can do this, and, when combined with Critical Path schedules the information becomes all the more accurate. But, since EVM/CPM systems are germane to we PM-types, none of the other specialties can come even close to the relevance and accuracy provided by these information streams.

Which brings us back to Lt. Commander Troi. If you have the ability to know that the Romulans are getting ready to double-cross you, step up and make that known, and sooner rather than later. Similarly, if you, as a PM-type (and member of GTIM Nation) know, based on your EVMS, which projects are doing okay, and which are failing to disclose probable overruns, step up and point it out, even if your assertions run counter to the other “experts” in the board room. Your cost and schedule performance information is accurate and relevant, theirs isn’t.

Otherwise, all of your PM expertise is analogous to sitting in a bridge chair, not connected to an actual station, and bleating “I sense they’re hiding something, Captain.”

As I wrap up GTIM’s take on COVID-19 Impact, one year later (ProjectManagement.com’s theme for March), I want to put a bow on the whole what-happened-and-how-do-we-move-on business, from a Project Management perspective (of course). A lot of how we get made whole on the project side of things is going to depend on what kind of contract your Team is working, and which kind of customer. As they (should) teach in risk management (no initial caps) school, the primary vehicle for managing your project’s risk is the contract vehicle itself. If you’re working a firm fixed price (FFP) contract, then your organization has signed on to accept all of the risk for the project, global pandemics (virtually always) included. If, on the other hand, your customer has signed on to a cost-plus contract (e.g., cost plus fixed fee [CPFF], cost plus award fee [CPAF]), then they have committed to sharing some of the risks involved in the project, and your job is now to articulate the best possible case for them to accept as much of that risk as possible. Add to this the PM’s ability to read their customers, and this is where things get tricky.

Just to be clear, nothing in this column should be taken as legal advice when making a force majeure claim. Those questions are best left to the contract administrators and lawyers, and I am neither. This blog is entirely from a Project Management point of view, and my own personal perspective at that. But it seems to me that PMs who have seen a sizable impact from COVID-19, and are working some form of cost-plus contract, could help themselves – and their Teams – by reading their customers’ mood or disposition as they advance their Baseline Change Requests (BCRs).

First off, was your Project’s performance harmed by the pandemic? A simple drop in Cost Performance or Schedule Performance Index (CPI/SPI) can’t be interpreted as ipso facto evidence of COVID-related damages. As I mentioned in an earlier blog, we’re looking at a series of potential scenarios, both with respect to changes in the performance indicators and the condition of the Project both prior to and after the lockdowns became prevalent.

The preliminary analysis has to do with the CPI and SPI behavior. Generally speaking,

• If both CPI and SPI are over 1.00, you are in great shape.
• If CPI is below 1.00, but SPI is over, that simply means you’re accomplishing the Project’s scope faster than originally baselined. If this is perceived as undesirable, simply tap the brakes.
• If SPI is below 1.00, but CPI is over, get on the gas. What you’re doing, you’re doing efficiently – you’re just not doing enough of it.
• Finally, if both CPI and SPI are below 1.00, you are in trouble.

Pretty basic, right? So now let’s overlay this onto the performance figures your Project has seen since February 2020 (the last “normal” month for most of us) and the present. This behavior can be binned, so:

If you do decide to generate a BCR under these scenarios, here’s what you can expect:

Of course, if your customer routinely engages in the actions listed in the Response column for Scenario 6, you won’t be able to glean any additional information on the status of your request for baseline adjustment. But, if that is the case, you probably have a lot of other problems anyway.