When I took a statistics class in college, the professor told us about the fact that there was a high correlation between the crime rate and the number of dentists in a neighborhood. When the number of dentists went up, so did the crime rate. The correlation was a provable fact.
After making this statement, the professor suggested that if you want to reduce crime, all you have to do is evict the dentists from the neighborhood. Doesn’t that conclusion fit the facts?
Of course, the point the professor was making was that just because one observes a high correlation does not mean that you understand the nature of the relationship. For example, you may not know which is the cause and which is the effect.
Or, in this case, the crime and dentists worked together because of a third factor—both criminals and dentists prefer nice neighborhoods. Dentists like them because they can afford them. Criminals like them because there is nice stuff to steal.
In other words, there is no direct connection between crime and dentists. If you eliminate the dentists, you still have nice neighborhoods, so the crime will stay. So don’t be hasty in your conclusions when you see a correlation.
Strategies are based on assumptions. The better one’s assumptions about the environment, the more likely one will have created a strategy relevant to the environment.
Today, we live in a world where big data and statistics are becoming a larger part of the decision-making process. Big computers can crunch all kinds of data and point out all sorts of correlations.
But just because we can crunch out ever larger lists of correlations does not necessarily mean we understand the environment better than before. Like in the example in the story, a factual correlation between crime and dentists can be misunderstood and lead to the wrong conclusion.
We can make the wrong assumption about the direction of causality (which is cause and which is effect). Or we may be missing out on a third factor which connects the other two (like the desirability of nice neighborhoods).
If we do not understand the underlying relationship in the correlation, we will make the wrong assumptions. And this will lead to having the wrong strategy.
The principle here is that it is wise to spend time critically examining our assumptions before plunging into strategic solutions. After all, a “great” strategy based on incorrect assumptions is really in the end a bad strategy, because it is irrelevant to the marketplace in which it will be implemented.
An ExampleThis point was made clear to me in an article I was recently reading about social media and shopping. The article talked about a correlation fact: customers who are heavily involved in a retailer’s social media spend a lot more money with that retailer than those who aren’t involved in that company’s social media. The article then concluded that based on this fact, retailers should put a lot of effort behind getting as many people involved in their social media as possible.
Now what assumption did that author make? Based on that conclusion, the author apparently assumed that social media activity leads to greater shopping. But is that really true?
What if the correlation works in the opposite direction? Doesn’t it make more sense that those who really want to spend a lot of money at a particular store would be the ones most likely to also like their social media?
So does more social media usage create heavier store shopping or does heavy store shopping create more social media usage? The answer to this question can make a big difference on what strategy one chooses.
If the author is correct, then one should spend all sorts of money to get as many people as possible involved in your social media, regardless of who they are and how you got them there. After all, the social media usage is supposed to magically convert them into heavy users, so it is worth the effort.
Another fact is that the most effective way to get people involved in social media is via “bribery.” In other words, if you offer money, prizes, coupons or contests (which are really nothing more than bribes), you will get more people to sign in and “like” you. So if you have an assumption like that author, you will endorse a strategy including lots of this kind of bribery.
But think about this for a minute. If the only way you can get someone into your social media is due to bribery, why do you think this will lead to a lot more loyalty and a lot more purchasing?
Now what if the other assumption is true, that heavy spending leads to heavier use of the brand’s social media? Then, one would look at their social media as targeted tool for communicating with some of their best customers. In this case, the bulk of the strategy might be pointed at trying to increase the amount of money these already loyal shoppers are spending with you. So instead of primarily trying to get more people to the site, you primarily try to extract money from those who naturally want to be there.
For example, you might come up with new offers of new products or services that would appeal to heavy users. Or maybe you offer them a loyalty program which rewards spending even more with your company (which, since they already like you, might be relatively cost effective).
If this second assumption is true, the tactics of the first assumption could be a disaster. For example, if you clog up the social media with a lot of people who really aren’t that interested in the brand, it might make the social media less inviting to your serious shoppers. This could chase away your best customers from the site and reduce the ability to have meaningful (and productive) conversations with your best customers.
Second, if bribery is what got them to the site, then significant levels of additional bribery are probably needed to convert these social visits into shopping. That could end up being a highly unprofitable loyalty program if you add up the bribes to get them to the site and then additional bribes to convert them into shoppers. At the very least, it would probably be less effective than a loyalty program only focused on people already favorably pre-disposed to your brand (where the needed bribes could potentially be far smaller, since they already have more loyalty towards you).
So, as you can see, assumptions can really impact strategic effectiveness, both positively and negatively.
So What Should We Do?This being the case, what should we do? First, become explicit in outlining your assumptions. Make sure you get those assumptions out in the open. Always ask yourself these questions: What assumption needs to true for this strategy to succeed? What assumption is embedded in my conclusion?
Second, take the time to really study and test the validity of those assumptions. Don’t rush to a conclusion merely because you see a correlation. Make sure that you test the directionality of the relationship or check to see if there is a third variable holding the correlation together. Do a test before you roll out your plan to see if the assumption holds true.
Third, consider the downside risks if your assumption is wrong. The greater the potential negative impact to a small error, the more you should study the validity of your assumptions.
Finally, don’t lull yourself into a false sense of confidence just because increased analytics has supplied you with a lot more correlations. You still have to figure out what it all means. More data may just lead to more misperceptions and more false assumptions. More data means more to analyze and study, not an excuse to plunge ahead faster.
If you want a strategy ideally suited to your environment, then you’d better have an accurate view of what that environment really is. Our point of view is based on the assumptions we make about how that environment works. Therefore, we need to take time to identify and test the assumptions inherent to our conclusions. Otherwise, our conclusions could be wrong and lead us to the wrong strategy.
Before you go around shouting for eviction of the dentists, think through the logic which lead you to that thought. Remember, a little time spent up front to fine tune the assumptions can save you a whole lot of grief later on.