THE STORY
On May 6, 2010, the Dow Jones Industrial Average dropped by nearly 1000 points in a matter of minutes. Companies like Procter & Gamble and 3M lost more than 30% of their value in through computer trades in just 15 minutes. Accenture went from computer trading at around $40 per share all the way down to a penny. There was nothing going on in the world that would create the sudden need for a market drop that extreme at precisely that moment.
A number of hypotheses have been suggested for why this sudden drop occurred. They include everything from a typing error by an analyst to cascading “Stop Loss” orders that were computer generated. The fact that a large portion of that drop came back before the day was over would seem to indicate to me that the original drop was not fully justified in any rational sense. This is further validated by the fact that the stock exchange cancelled many of those trades during that vulnerable time period. Apparently, they didn’t think those trades made sense, either.
Although we may never know exactly what triggered the mess, one thing we do know is this: The mess was compounded by computers blindly doing something which they were programmed to do, regardless of whether or not it was logical or sensible at that moment. Human logic was bypassed as the computers coldly and uncaringly executed orders that were triggered by programming algorithms put in place a long time ago. And because computers can do these calculations so quickly, they can distort markets well before human logic and common sense can intervene.
It seems to me that the more we rely on so-called “rational” computers to do stock trading, the more “irrational” the market swings become.
THE ANALOGY
There has been quite a bit written recently on the decision-making biases of humans. It has cropped up on the web in places like the McKinsey & Co. site, the Booz & Co. site, the Harvard Business Review Blog site and numerous other places.
Many of these biases have been well researched and have become a part of “common knowledge.” As a result of these biases, many “experts” are advising leaders to set aside their gut instincts and instead just load up on numbers…and then let the numbers make the decision.
The assumption in many of these materials is that if we get rid of human biases and just rely on the numbers, we will get better decisions. Unfortunately, when I look at things like the stock market crash of May 6th, I get a little nervous in abandoning human intuition and relying 100% on numbers and computers. So-called “blind” decisions made by computers on that day don’t appear to have been logical or sensible. Why should I abandon my gut intuition to cold mechanics that can bring down the stock market? Couldn’t it do the same to my business?
The crux of the problem, as I see it, is the false assumption that just because you eliminate human biases, you have eliminated all biases. My contention is that numbers can be just as biased as humans. If you fail to take into account the inherent biases in numbers, you will be driven astray as much as you would be by human biases. In fact, I believe that the biases inherent to numbers are even more dangerous that human biases, because numbers cannot step back and ponder about their implications as humans can.
THE PRINCIPLE
The principle here is that numbers should not be accepted at face value any more than gut intuition. Both have biases that need to be brought out into the open. Until you deal with these biases, you will not be able to see reality. And until you see reality, you cannot make the best decisions.
Since there has already been a lot written on human biases, I will shift the attention of the remainder of this blog to some biases inherent to numbers.
1. Bias of Backwards
Most numbers relate to the past. They tell you about history...what has already happened. Therefore, the numbers are biased by that historical context.
The problem is that your decision-making is about the future. There is no guarantee that the context of the past will continue into the future. Therefore, numbers of the past may not accurately provide guidance for the future. As Warren Buffett put it, “If past history was all there was to the game, the richest people would be librarians.”
For example, consumer data collected prior to the tragedy of 9-11 (or the great recession we just had) may be worthless in projecting behavior after those events. As another example, consumer media behavior data gathered prior to the invention of the next form of media (radio, TV, vcr, internet, ipad, etc.) will not automatically give 100% accuracy to how behavior will change when the new medium appears. Looking at Baby Boomer behavior in their 30s may not give accurate insight into how Millenials will act in their 30s. Or for that matter, looking at how Millenials act in their 20s may not give accurate insight into how Millenials will act in their 30s, either. The contexts are different, so the results may also be different. It takes human intervention to make sense of the changes in context.
2. Bias of Availability
When gathering numbers, one is limited by the numbers available to gather. When looking at the pile of numbers you have gathered, the larger piles are not necessarily more important or more meaningful. There are just more of them available to gather. Just because a number is available does not make it relevant.
The problem is that a lot of strategic decision-making is about finding new, untaken white spaces in the environment—sometimes referred to as Blue Oceans. These new opportunities typically come about by reorganizing things in a new way—looking at things as they have not been looked at before. Almost by definition, if you are gathering numbers that are already available, they are biased by the conventional way of categorizing and organizing numbers.
For example, governments tend to categorize economic data based on the conventional, historical ways in which industries were classified. All the numbers neatly fit into the black spaces (which is why they are black). There aren’t any available numbers in the white spaces, because that’s not how governments deal with numbers. If you are planning to reinvent the marketplace and create new industries (or significantly redefine the borders of old industries), the available numbers will be distorted and not give a good picture of that redefined world. They may even keep you from finding that Blue Ocean, because all the numbers are allocated to black spaces, making it appear that the potential is empty outside those black spaces.
3. Bias of Source
Numbers come from someplace. People compiled those numbers, and the biases of those who created the numbers often distort these numbers. Just look at how different political parties can look at the same world and find numbers to support their radically different positions. So even if you do not inject your bias into the equation, the numbers you have may be full of biases from the people at the source of those numbers.
This bias from the source doesn’t have to even be intentional. For example, if you are trying to gather numbers about the future, there are not many options. Either you gather opinions from “experts,” opinions from consumer research, or results from computer models. Experts can have all kinds of biases that creep into their opinions, intentional or not. Scientists even have a name for this, called “Expert Bias.” Computer models are based on underlying assumptions placed there by people. There can be lots of biases in those assumptions.
Finally, consumers are notorious for their generosity in giving opinions about how they might act in the future, but even they do not really know how they will behave in a new environment. Their guesses often are quite different from what they eventually do. It’s not that they intentionally lie…they are just bad guessers.
4. Bias of Bigness
Numbers are often gathered at the macro level, yet consumers behave at the micro level. Macro level numbers have a bias towards bigness which distorts our conclusions about micro level behavior.
Take, for example, economic stimulus plans. A government might say it is pouring millions, or maybe even billions of dollars into the marketplace. Once could look at all that money at the macro level and conclude that surely customers would be stimulated to spend more on big-ticket items with that much money flowing towards them. However, if you look at the numbers at the micro level, you may see that the stimulus package results in an extra five dollars per household per paycheck. How many people are going to radically change their big-ticket purchases based on a five dollar increase in their paycheck? Heck, they might not even notice the five dollar bump.
In an earlier blog, we talked about how macro numbers tend to portray average behavior, whereas if you get down to the micro level, you can see all the extreme behaviors which are averaged out of the big number. By looking at big roll-up numbers, you may not see all the great niche potential hidden in those averages.
5. Bias of Unusual Event
Numbers gathered for a period of time are biased by the quirky little events occurring at that time. Outcomes would be different depending on the presence or absence of those events. For example, the green movement was on a particular trajectory. Then, Wal-Mart decided that it was going to take green issues very seriously. That one change in circumstances radically changed the trajectory of the green movement, because of the power and influence of Wal-Mart.
Consider that the entire evolution of the consumer technology industry has been influenced by the fact that Steve Jobs is in the business. Had he not been a part of Apple, the entire industry probably would have evolved differently. Have you factored into your numbers the possibility of a “Steve Jobs” person emerging who can change the course of history or the possibility that a powerful force like a Wal-Mart may change its direction? Or what if you decide to inject yourself into your industry the way Steve Jobs did or change directions like Wal-Mart did? You will not see this in the numbers, since those numbers did not include such a circumstance
When computer models are built to predict the future, the assumption is that all of the relevant events are included within the model. As the stock market crash of May 6th showed us, new events may turn up which were not accounted for in the model, causing the models to fail.
SUMMARY
We have been told not to rely on our gut instincts at face value, because they are biased. Similarly, I believe that we should not just rely on numbers at face value, because they are also biased. Instead, we need to rely on both our gut instincts and numbers, keeping in mind the biases inherent to both.
FINAL THOUGHTS
To borrow another quote from Warren Buffett, “Never invest in a business you cannot understand.” Just because you have a pile of numbers doesn’t mean you understand what is going on now (and what could go on in the future). You need to comprehend what is behind the numbers. This takes insight, something which transcends mere data and must incorporate human intuition and imagination.
No comments:
Post a Comment