Let me start with some of my own shopping behavior. I occasionally do grocery shopping for the family My store of choice is Schnucks, mainly because I understand their layout. One curiosity I've noticed is that in the aisle with cocktail olives (for martinis) you can also find peanut butter. I consume both of those items but in my view of the world, they don't go together and I don't consumr them at the same time. Cocktail peanuts do go with cocktail olives (the appropriate munchie to accompany the martini) but you won't find cocktail peanuts in this aisle. They are in the aisle with the potato chips and other snacks. Cocktail olives also go with whatever liquor is used to make the martini. So you might envision the olives in that section. But you'd be wrong. Some people do put cocktail olives into salads, so conceivably you'd find them in that part of the store, but that too is not the case. For me its a mystery why they are with peanut butter in the same aisle, one I'm not likely to solve soon.
Stores like Schucks now generate a huge amount of data via the checkout process. Every item has a bar code on it. Every item gets scanned at checkout. This is an incredible boon, not just to getting through the line quicker, but also for inventory management. The store gets a much better sense of what merchandise it is moving and what remains on the shelves. So for restocking standard items, this is fantastic. Does it help the store think through where items should be located? I know from the little I've read about marketing in grocery stores, that the buyer should see fresh produce when he first walks in and that milk and fresh meats should be in the back of the store. There are known buyer behaviors that support these product location choices. What about other such choices? Are those choices driven by data - let the numbers do the talking? Are they driven by some conception of the shopper that may or may not be supported by the data? Or is it something else?
More generally, with this shopping as only one example of the bigger picture, are we talking only about making small tweaks in an already well defined model? If we are, does that really require human discretion, or can it all be computerized and run by algorithm? If it can be done by computer, it would seem more information is better, in contrast to what the book says about information efficiency. With a higher dimensionality of information, one can be more precise about the actions that need to be taken. That is one way to think of what's going on.
Now I want to give a completely different perspective. For this one, you can imagine you are a medical doctor who has a patient come in to see you with a variety of complaints. One of the things you need to determine is whether it's all minor stuff, some topical treatments and perhaps an inexpensive prescription will do the trick, or if it is a major illness, requiring either substantial surgery or a very invasive treatment regimen. Further, if it is the latter, you may have to come to a decision rather quickly, because long delay can cause the situation to deteriorate further. In this same manner you can imagine you are a different type of doctor, the Chairman of the Federal Reserve, with the patient being the entire U.S. economy. You need to determine again whether the economy is basically healthy or if there are some serious underlying issues that need fixing.
Suppose in particular that you are Alan Greenspan and it is back in late 2004 or early 2005. If you thought the economy had a serious illness in it back then, could you have prevented the meltdown from happening a couple of years later?
This diagnostician has a huge amount of information at his fingertips. Much of it is digital in nature but - financial information about prices and quantities regarding bank balance sheets. But you also have a huge amount of analog information - reflecting the attitudes of the circle you trust and the information that is shared within that circle.
In retrospect, where vision is always 20-20, Greenspan blew it completely on the economy. According to the macro economist John Taylor of Stanford - the Fed held down interest rates for too long. That induced the bubble. The Fed also had regulatory authority it didn't exercise about the subprime market. Greenspan didn't see the need to change what he was doing, because he thought it was working well.
It turns out that on making the judgment of how the patient is doing, a big part of the issue is the framework that is used to process the available information. In Greenspan's framework (you can call it a mental model), the higher ups in the big financial houses had a strong self-preservation instinct, not just for themselves but also for the organizations where they worked. This self-preservation instinct meant these people would self-regulate their own organizations to keep them healthy. Greenspan therefore couldn't imagine taking on bad debt for near term gain. He therefore so no reason to sound the alarms.
I call this type of behavior believing in myth. I think we all do this to some extent. The issue arises when the myth comes into conflict with some available evidence. Then something must go. When it is the myth that goes, I call that behavior being empirical. One then needs to come up with a different framework to explain what is going on. You'd like to believe that science works that way. However, it sometimes happens that what gives is the evidence, and the myth survives. In this case, you have to wait for a catastrophe to see what is happening, and then you can go back to the evidence earlier and note you might have seen it earlier too, had you not been blocked by an erroneous framework.
The Lessons from Tal Afar has this aspect as well. The crucial question there was whether there was an insurgency or not. One gets a completely different military strategy when there is no insurgency than when there is an insurgency. The latter seemed to produce much better results.
It may be that the big thing senior management does is communicate a framework on which to hang all other decisions. Then information flow from the spokes back to the center can be seen in part as a way to support the effectiveness of the framework or to demonstrate the flaws in it. If this is right, then a further management task is to make adaptations in the framework or wholesale changes in it, when those flaws become apparent.
Armed with a reasonable framework, then one should be able to determine what resources are necessary and what actions need to be taken to be successful. There may have to be lots of minor tweaks to the recommendation, to match the situation on the ground, which is likely in flux. Those tweak things are not the focus of the senior management. It is middle management that does the tweaking, or the task is automated.
Does the information efficiency concept in the book survive if we're talking about management by selection of a framework? Perhaps it does, but here the arena is the verification part. If the complexity is too great, one can't tell whether the framework is useful or not. In a simpler environment one can tell. Information efficiency then says to get the simplest framework possible that will really test the whether the framework is applicable.
No comments:
Post a Comment