At the tender age of 17 I sold my soul to a 1979 MG Midget. Or, at least I sold my body into indentured servitude to it. I spent my entire earnings from two prior summer jobs and my part time job after school to acquire this little gem. Fire engine red. Black racing stripes. Mag wheels. And, yes, a sixty-five bhp engine. As cool as that sounds, and as anyone who has ever owned a British Leyland product already knows, I spent a lot of time tinkering with the car to keep it running, often at the most inconvenient of times.
Doe-eyed brunette: “Why are we stopping here?”
Me: “Uh, I have to tinker with the engine. Seriously.”
I obsessively measured the performance of this car by the fuel mileage. With each fill-up I recorded the gallons and the change in the odometer reading in a little notebook I kept in the glove compartment. By this, I kept track of my fuel mileage between fill-ups and a running average over the month, each to unwarranted precision. My vigilance didn’t tell me much, just enough to know when the British Leyland works of the car begged for yet another tune up–my never ending slavery under the oak in my parent’s back yard. Considering the cost of gasoline in 1985 (~$0.90/gal), my obsession may have saved me $10 over the course of a year.
Fast forward to the 2000s. Having redeemed by indentured obligations, I now own a much more modern car with an advanced onboard computer. It’s so advanced that I have absolutely no idea how to tinker with this car. In fact, I really do not need to except to change the occasional burnt out taillight or turn signal bulb. However, my obsession with fuel mileage has intensified because I can get readings not just on average fuel mileage but also on instantaneous fuel mileage. I possess insights now that I never had before. The advantage of having instantaneous fuel mileage readings is that I can immediately observe which driving behaviors correlate with the best fuel mileage and take immediate steps to coax my car toward better efficiency.
Please don't misunderstand my point. Knowing the average is beneficial. By monitoring its trend I can make inferences about the health of my car. In fact, a dramatic drop in efficiency a few weeks ago alerted me to a problem with my fuel system (or so my mechanic told me after I gave him $1100 to ask my onboard computer what the problem was). However, aside from taking immediate action in response to rapid changes, it's difficult to know from watching an average how to manage the average. I might observe shifts and fluctuations, but I don't know which factors are actually driving the average since an average is an aggregate effect of all the influencing factors accumulating over time. Getting nearly instantaneous feedback allows me to make associations with concomitant events that likely do influence the fuel efficiency of my car. I say “likely” because there is always the possibility that I assign cause cum hoc ergo propter hoc.
Another benefit arises from getting instantaneous feedback and observing causes. I don't just make near term corrections as I observe the need. I learn the causes of improved efficiency and apply these insights to improve the likelihood that I achieve my goal proactively. For example, I don't accelerate as quickly as I did when I was 17. I also coast downhill a lot, letting the momentum of my car do more for me than the engine. I estimate that driving this way improves my average efficiency by about 1 mile per gallon, putting about $103* back in Starbuck's cash register this year alone.
I hope it's obvious that I'm not just talking about how to maximize automobile fuel efficiency here. The above discussion is an analogy for risk analysis and management. Risk analysis is about discovering the causes of variance from a desired goal and the causes’ relative strength on that variance. Risk management is about controlling those causes of variance. Responding to an average is usually not informative enough or timely enough to maximize the likelihood of achieving an objective, because, as I pointed out earlier, an average is an aggregate of information over time, sometimes a very long time during which specific causes maybe forgotten.
Why is this on my mind? It happened to be the topic of discussion with a client a few weeks ago as we thought about managing the cost and schedule of a large, complex, capitally intense construction project. The client expressed frustration that past projects had been poorly planned and managed because many of the supporting analyses and risk mitigation plans were based on traditional “averages” (which we all know were really biased assumptions and not even proper averages). The above analogy helped him see why we need to base our analysis in a form that helps us anticipate causes of variance and their relative strength on the desired outcomes of “on time” and “within budget.”
(Next month, I’m going to discuss some useful ways Analytica can be employed to estimate causes of variance and their strength.)
*15,000 miles x (1/22mpg – 1/23mpg) x $3.50/gal