134
- THREE -
CAUSES
"We measure and reward performance by the quantiry of steel produced. "
What do we want to accomplish? It's hard to achieve a result if we don't understand what causes the result to happen. In order to solve problems or achieve goals, we must first understand what causes the result we want to accomplish. Start with examining what factors make up the system and how they connect. Then, define the key factors that determine outcome.
If a business measures performance by the amount of steel produced, they will
get a lot of steel produced. But the amount of produced kilo steel is only one part of the equation. It's better to ask: What is the equation that achieves what we want to accomplish? What factors cause what we want to achieve? Under what circumstances? What causes business value? Do we have the factors needed? What must be changed in the equation to achieve what we want? Have we thought through what other effects our actions may have?
Large effects
A bird flies into the engine of an aircraft and disaster strikes.
We believe that cause resembles its effect - for example, that large or important effects must have large causes or that complicated outcomes have complicated underlying reasons. But the size of an effect may not be proportional to its cause. Small things may break a large system. In 1988, 35 people died and 113 were injured when a London-bound commuter train crashed into the back of a stationary train. The accident was caused by faulty wiring work. A mechanic worked more than 12 hours with only a 5-minute break and forgot to remove a small wire from an old switch when installing a new rail signaling system.
On September 23, 1999, the Mars Climate Orbiter spacecraft disappeared
What happened? The core cause was the failed conversion of English units (pounds) to metric units (newton). The manufacturer measured the force of small control thrusters in pounds but the space officials expected newtons. A misunderstanding that sent the spacecraft about 56 miles too close to Mars, making it disappear into the Martian atmosphere. A simple mistake caused the loss of a $125 million spacecraft.
135
Washing hands save lives.
Dirty hands can cause major health problems. Among bacteria that reproduce by dividing themselves every half hour, a single one can generate about 17 million offspring within 12 hours. Records at the Center for Disease Control and Prevention (CDC) show that every year close to 2 million people (out of 35 million admissions annually) in the U.S. pick up an infection in the hospital. Of those people, about 90,000 die as a result of their infection. And the major reason is bad hygiene like unsanitary conditions, germ-laden instruments and unwashed hands. CDC and the U.S. Department and Health and Human Services estimate that strict adherence to clean-hand policies alone could save the lives of 20,000 patients. That washing hands saved lives was discovered in 1847 by the Hungarian gynecologist Ignaz Philipp Semmelweis. When working at the Maternity Department of the Vienna Hospital, he observed that women delivered by medical students had high mortality rates while those delivered by midwife trainees had low rates of childbed fever. The difference? The medical students had made pathological dissections, or had come into contact with dead bodies before the examination of the women. Semmelweis introduced the practice of washing hands in a solution of chlorinated lime before every examination and mortality
rates plummeted.
Random events
'.11. fire in our supplier's factory caused delivery problems of components. We lost market share to competition, which resulted in huge losses. This outcome was unforeseeable since we couldn't have predicted the fire in advance. "
When bad things happen, we try to find causal explanations or something to blame. The more unexpected or negative we find an event, the more likely we are to look for explanations. We underestimate the influence of randomness.
Here we define a random event as an event that can't be predicted because of a lack of knowledge. An event is random when we don't have enough information to determine its outcome in advance.
Acting on symptoms
Take away the cause, and the effect ceases.
- Miguel De Cervantes (from Don Quixote)
Sometimes we mistake an effect for its cause. There is a story about a man that was walking by a river when suddenly a screaming girl floated by. The man jumped in the river and saved her. After five minutes, another screaming girl
136
floated by. He jumped in again and saved the girl. The same thing happened over and over again. The problem was a little further up the river. There was a man throwing girls from a bridge. Our hero solved the symptoms but not the cause of the problem.
"We have tons of problems. We are losing customers, we can't deliver on time, our inventory system doesn't work. "
What is the core cause of these problems? Many times when we have a lot of problems, there may be one common reason for them all. When dealing with problems we must focus on what we want to achieve and make sure that we address the underlying cause and not act on symptoms that may look like causes. Maybe the symptoms were due to wrong policies or measuring instruments or goals, etc.
Multiple causes
We attribute an outcome to a single cause when there are multiple causes. We assume that A causes B but A may not be the only thing that causes B. There may be many causes for a given effect.
For example, behavior is determined by a multitude of psychological and non psychological factors, individual characteristics and the given situation. Sometimes these factors interact and reinforce each other. When explaining behavior, think in terms of multiple causes. For example, Charles Munger says on the Milgram experiment discussed in Part Two (where a group of research subjects delivered electric shocks simply because they were told to do so by an authoritative figure):
For years it was in the psychology books as a demonstration of authority- how authority could be used to persuade people to do awful things. Of course, that's mere first conclusion bias. That's not the complete and correct explanation. Authority is part of it. However, there was also quite a few other psychological principles [consistency, contrast, reason-respecting], all operating in the same direction, which achieved that lollapalooza effect precisely because they acted in combination toward the same end.
Mistaking correlation for cause
Correlation means a relationship or association between two or more variables. We tend to assume that when two things happen together, that one causes the other. That a change in one variable is strongly correlated or followed by a change in another doesn't automatically mean that one causes the other. Some third factor may cause them both. Assume we detect a high correlation between money
137
and happiness. But that doesn't tell us if money causes happiness, if happiness causes money, or if some third factor causes them both.
An example of statistical misuse is from 1992 when it was reported that 28 teenagers who frequently played the game of "Dungeons and Dragons" (D&D) committed suicide. What conclusion should be drawn? Is there a link between teenage suicide and the game?
The American mathematician John Allen Paulos tells us in Innumeracy to put this statement into the right perspective by considering two more facts. The game sold millions of copies and about 3 million teenagers played it. In that age group the annual suicide rate is about 12 people per 100,000. This means chat we can expect 360 D&D playing teenagers (12/100,000 x 3 million players) to commit suicide.
Alternative explanations
''Mary has a fever and therefore she has a cold. "
There may be many explanations for a given outcome. But we often jump to conclusions and fail to consider alternative explanations. A given effect may be consistent with a range of causes and don't help us find the core cause of a problem.
That Mary has a fever tells us that she is not well but it doesn't tell us why she is ill. Fever can occur from a range of diseases. Ask: What else can explain this ou
tcome?
Selective data and appropriate comparisons
There is nothing more deceptive than an obvious fact.
Sherlock Holmes (Arthur Conan Doyle, The Boscombe Valley Mystery)
We identify the wrong cause because it seems the obvious one based on a single observed effect. As Bertrand Russell says: "Obviousness is always the enemy to correctness."
When someone remarked to the French writer Voltaire, "Life is hard," he retorted, "Compared to what?" We tend to ignore alternatives, and therefore we fail to make appropriate comparisons. Often we only consider information or evidence that is presented or available and don't consider that information may be missing.
"By studying successful businesses, I have found that their secret to success is focus," said the CEO ofTransCorp.
But many failed businesses also had focus. Is success, then, due to strong culture, visionary leaders, or something else? Maybe failed enterprises also had strong
138
cultures and visionary leaders. To understand what qualities cause success, we need to compare business successes with failures. We must include companies that started with the same quality but failed. The problem though, is that the data for failures often disappears. Furthermore, what works in one situation may fail in another.
By interviewing survivors of fatal airline accidents, a 1986 Discover article presented some tips on how we could improve our chances of surviving plane crashes. But as noted by Psychology Professor Robyn Dawes, there was no evaluation of what percent of all passengers did what the survivors did. Maybe those who died also had done the same. Of course, we can't interview the passengers who didn't survive.
Often the same attributes are used to explain both success and failure. For example, when a company is successful, the media say that it is because of its focus and great leadership. When performance goes down, they report that the company became less focused and its leadership deteriorated. While there may be evidence that the company is less focused or has worse leadership than before, it could just as well be that the competition has gotten better. Business performance is relative, not absolute. The performance of a business is always influenced by what its competitors do. Even if a company gets better in many areas, market share and profits may decline if the competition does things even better. And even if a company gets worse in many areas, market share and profits may increase if competition does things even worse.
"Look at where the bullet holes are and put extra armor every place else. "
During World War II, the statistician Abraham Wald tried to determine where one should add extra armor to airplanes. Based on the patterns of bullet holes in returning airplanes, he suggested that the parts not hit should be protected with extra armor. How could he reach that conclusion? Because he also considered planes that didn't return. Assume that all planes had been hit more or less uniformly. Some planes hit in marked areas were still able to return. This means that planes that didn't return were most likely hit somewhere else - in unmarked places. These were the areas that needed more armor.
"The drug obviously worked since I used it and got better. "
But the same outcome could have happened without taking the drug. We need to consider both confirming and disconfirming evidence. Ask: What is the frequency of supporting cases compared to disconfirming cases? What is the relative frequency of this condition or disease in the population?
For example, when studying the effectiveness of a new treatment, people often
139
ignore the outcome of a non-treatment, and focus only on a large number of successful treatments, and therefore conclude that the treatment is effective. But we also have to look at the number of people who a) use the treatment and don't improve, b) don't use the treatment and improve, and c) don't use the treatment and don't improve.
In one study people were asked to judge the effectiveness of a treatment based on the following data from an experiment:
Treatment No treatment
Improvement 200 people
50
No improvement 75
15
Most subjects believed the treatment was effective. But it is rather ineffective. Why? We need to compare the outcome for the no treatment group with the outcome for the treatment group. 50 of 65 people (50+15) or about 77%, improved without any treatment versus 200 of 275 (200+75) people or about 73% who improved with the treatment.
Psychology Professor Thomas Gilovich says in How We Know What Isn't So: "With the body so effective in healing itself, many who seek medical assistance will experience a positive outcome even if the doctor does nothing beneficial. Thus, even a worthless treatment can appear effective when the base-rate of success is so high."
What caused customer complaints at TransCorp?
We look for obvious causes and look at the conditions and behavior that were present in the particular situation after we know what happened. But we should observe what is normally happening. Most outcomes are context-dependent.
Instead of concentrating on conditions, and behavior preceding the customer complaint, TransCorp should examine both sales without complaints and with complaints and ask: What were the underlying conditions and behavior when customers didn't complain? Are there differences that can explain the complaints? For similar situations, ask: Compared with what? Compare negative outcomes with positive. What factors differ? What were the underlying conditions and behavior where something bad happened compared to the conditions and behavior when things worked? What differentiates the situations? Don't draw
conclusions from what may have been a unique or random event.
140
- FOUR -
NUMBERS AND THEIR MEANING
Use basic math to count, quantify, and understand relationships
john bought all the stock he could since it was priced at only $1.
Something is only cheap or expensive in relation to something else. If an investor refers to a stock as "cheap," what is it cheap in relation to? 100,000 shares of stock priced at $1 have the same market valuation as 1,000 shares priced at $100. The relevant question is: What is the company worth in relation to its price?
TransCorp made a "huge" $1 billion in profits.
Words like "big" or "small" have no meaning in themselves. A number has only a size in relation to another number. $1 billion says nothing about economic performance unless we compare it with how much capital was needed to generate it. What ifTransCorp needed $100 billion in equity and debt to run the business? That's only a 1% return.
"Research shows that the new drug reduces the risk ofgetting the disease by 25%." What benefits can be expected by taking the new drug? Assume statistics show the following: Without the drug 20 people in 1,000 get the disease. By taking the drug, the figure drops to 15 people in 1,000. How efficient is the drug if we look at the absolute number of people saved from the disease? The reduction in absolute risk is 5 people in 1,000 (20-15) or 0.5%. The reduction in relative risk or the relative number of people saved from the disease is 25% (5/20). A 25% reduction only means something if many people are saved.
How many people need to take the drug in order to save one individual from the disease? Since 5 people in 1,000 (or 1 in 200) are saved from the disease by taking the drug, only one person out of 200 that take the drug will be saved. For the other 199 people the drug does not have any positive effect.
How do we know that research findings are true? Surveys of epidemiological studies (studies of factors affecting health and illness) show that many researchers' claims don't replicate in medical trials because of bias and random error. For example, two refuted claims are that aspirin is highly protective against heart disease in both men and women and that vitamin E reduces the
141
chance of cardiovascular mortality. Studies by Hygiene and Epidemiology Professor John Ioannidis show that the smaller the studies; the smaller the effect sizes; the hotter the field; the greater the interest; the larger the databases
; and the greater the flexibility in analyses, the higher the chance that a research claim is false. True replication is the cornerstone of science, and assuming that biases are weeded out, the more researchers reproduce a finding, the better chance it has of being true.
What does it mean that our sun is 93 million miles away?
Often we need to translate a number to something more understandable. Light travels at an average speed of 186,281 miles per second. This means that it takes sunbeams about 8 minutes to reach the earth. The star Alpha Centauri is 4.35 light-years away. A light-year is the distance light travels in one year. Even iflight year measures distance, it implies time. If we were to ride on a beam of light, it would take us 4.35 years to reach our closest star. When we look up in the sky we see the past - the star as it appeared 4.35 years ago.
Always look at what numbers mean. For example, ask: Does the magnitude make sense? In relation to what? Also, think about what counts - not how it is counted. There was a sign hanging in the physicist Albert Einstein's office at Princeton that said: "Not everything that counts can be counted, and not everything that can be counted counts."
Seeking Wisdom Page 20