Monday, November 23, 2009

Bill Belichick is the model for management success

This is a great point - not about football but about the typical mid-level manager. Their incentives are not always about maximizing return or minimizing risk to the company. Often, their incentives are to minimize the probability that something they do will get noticed in a bad way by upper management. Just keeping their heads down is the best path forward. Unfortunately, this aggregates across the company into performance that is much lower than it could be. Companies need to encourage more Bill Belichicks.
Interesting post by Daniel Isenberg at Harvard Business Review. But I have to disagree in part.

There is a lot of evidence that incremental innovations yield more over the long term than breakthrough innovations, in part because there are so many more of them and that they are cheaper to develop and commercialize. BUT there are two big BUTs.

1. In the longer term, it is the breakthrough innovations that lead t...o real improvements in quality of life. So if we want to make a difference in the world, we need to have breakthrough innovations.

2. It is the pursuit of breakthrough innovations that create talented innovators. These people may be necessary to keep the stream of incremental innovators going. So if we all shift to "minnovation" as this HBS blogger suggests, we may reduce our innovation capability in the long run.

Thursday, November 19, 2009

Breast Cancer Screening

I haven't seen the actual data, so I am not going to opine on whether women from age 40-50 should get mammograms or not. But just in the one day that this discussion has been going on, I have seen so many classic decision making biases I can't resist discussing those.

1.Comparing statistical evidence and anecdotal evidence. No matter what the debate is about, there is always a great story that illustrates the benefits of one side or another. This is true whether your side has any merit or not. Anecdotes are a great political tactic, but should never be used to set policy. And yet, I have heard so many stories of the "42 year old woman named XXXX whose life was saved because she had a mammogram. Not only is she thankful for the screening, but so are her husband, teenage daughter, . . . ." Of course that will happen when millions of women are screened. But it is irrelevant to the policy decision.

2. Ignoring the costs of false positives. For every cancer that is found (true positive) there are millions of false positives. These false positives are real women too. They go through the fear that they have cancer, the pain of a biopsy, the hassle of the procedure, etc. It's not just about rationing care to lower costs. There are emotional costs as well. And there are many more false positives than there are true positives (see next comment).

3. Ignoring base rates: I have also heard the testimonials of women who say "Yes, I was afraid for a few days and had to undergo the pain and hassle of a biopsy. But it was worth it to find my cancer. So the cost benefit analysis is clearly in favor of screening." But what this argument fails to consider is that millions of women go through the fear, pain, and hassle for every cancer that is found. And in many cases, the cancer would have been found and treated even without the mammogram so even the true positives are not really true positives.

4. Validity bias. The people being quoted or interviewed vary tremendously on whether they are experts on the topic. For some reason, a cancer survivor is seen as an expert on breast cancer diagnosis or statistical analysis. Sorry, but having cancer does not make you an expert. Not even on the pain and suffering part because each woman's experience is different and your pain, while real, may not be typical.

5. Sample size bias. Also in these interviews, people toss in the results of studies. But if one study looked at thousands of women and another looked at dozens, they are not comparable. And yet, the discussion doesn't account for this.

6. Availability bias. It's much easier to think of the extreme cases (where a woman's life was saved by screening) than the more typical cases (false positives, minor cancers that would have been caught anyway, etc.). So the debate focuses on these salient stories instead of the real evidence.

7. Confirmation bias. Once a person decides which side they are on, they completely ignore all of the evidence to the contrary, even when it is being presented to them directly in a one-on-one discussion. It's amazing how good we are at this.

There are more too. But this is enough for today. Can we perhaps focus more on intelligent policy development instead of emotion-based policy development so we can actually create useful and effective policies? Anyone? Anyone?

Friday, November 13, 2009

Viva La Difference Part Deux

Ok, here is the second part I promised. Should we take these differences into account in our public policy? It's one thing for free citizens to make free decisions about how to design a product or what products to sell or buy. But should people be forced to do so? The libertarian in me says no. But of course we should think about it before jumping to conclusions.

Let's take public school policy. The government has a stake in this because a stronger educational system would lead to stronger economic growth and a higher standard of living for us all. So lets say that boys learn better with blue textbooks and girls learn better with green textbooks. Do we mandate to textbook companies that they need to spend the extra money to create two colors of textbooks, and then mandate that school systems spend more to buy them (as the increased costs would invariably be passed on)?

One example of gender-based policy is the famous Title IX which mandates that schools need to have equal numbers of male and female athletes, regardless of the interest of students, fans, or revenue. This means if a school has 100 male football players, they need to support 3 or 4 female sports to make up the difference - since most sports have 25-30 players.

If men respond better to a blood pressure drug than women do, should we mandate the the drug company spends extra money finding a similarly effective drug for women?

If you are answering yes to some of these and no to others, I will repeat a comment from a previous post. Do you have some logical basis for the difference?

Thursday, November 12, 2009

Viva La Difference!

I have been seeing a lot of research on gender differences recently. Not just in human factors, but in areas as diverse as education, economics, design, finance, nutrition, psychology, happiness, etc etc etc. There are two important overarching issues that this entire line of research brings up. Do we design differently for males and for females? And should government create regulations that compel systems to be designed differently for males and females when there is a public interest at stake. I will cover the first in this post and the second in another post.

First, should system designers create different systems for each gender? When it comes to simple things like apparel, it’s pretty obvious. Men and women have different anthropometry and different style preferences and there is no conflict. We have always had different clothing choices for men and women. Good thing too. I just don’t look very good in a dress and I can’t imagine trying to walk in heels.

But let’s think more broadly about this issue. There are biological differences that lead to different nutritional needs. Should our breakfast cereals be fortified differently? Different meal choices? And with health care we need different drugs. How far do we go with this? We already have drugs specifically targeting male and female diseases (breast cancer v prostate cancer being an obvious one), but what about making different versions of blood pressure medicine? How about different versions of Amazon.com to match different web navigation styles or background color preferences? When you log in (or using cookies) you would be automatically directed to the right version.

It gets harder when we talk about education. Boys and girls seem to learn better in single gender classrooms, but then their socialization might lag. A hybrid school may be ideal. More research would be needed to find out the best combination. But is this a direction we should pursue? What if boys learn better in an all boy classroom but girls learn better when it's 50/50? What would we do then?

I’m not really sure where I am going with this train of thought. Since most design stems from business needs, I suppose the choice comes down to whether the differences create enough customer demand that people will pay more for gender-customized products in each market. They would have to be willing to pay more to cover the added R&D, design, and manufacturing costs required to make two (or more) different sets of products in each category. For apparel, obviously the decision was made generations ago. These questions are just now being asked in these other industries. I am curious how far it will go.