More on risk measurement– recency bias

Measuring risk is a critical, yet impossible task. How can we act effectively without understanding the risk? How can we see the future? Or, what is the right balance between educated estimates of risk and potential benefit? These questions are important in many aspects of life, not least pricing. A lot of companies make pricing decisions without a good understanding, or even an attempt to measure risks. And, as we’ve seen recently, even businesses who are supposed to be really good at measuring risk– banks– can screw up terribly. Now your business needs to weather a bad economy, while subsidizing those who are supposed to be experts at measuring risk.

I wrote a post a few of weeks ago about the risks in measuring risk. Here is another article on the pitfalls of simplistic risk measurement, focusing on recency bias. Recency bias– the natural tendency to weight recent events more heavily than earlier events that are just as statistically relevant– is a common trap in risk assessment. The article notes that if you took a simple view of the risk of recession in any economic quarter, your time horizon has a huge effect on your results. If you only look over the past 20 years, the threat of a crash appears very slim– once every 624 years! But if you go back to WWII, the same measurement leads you to expect a crash about every 8 years. What’s really interesting is if you graph the odds of a crash over any 20 year period, the results change dramatically depending on which 20 years you pick.

Here is the graph from the article (clicking will take you to the article):

Note how the relative stability of the last 20 years has caused a lot of people to “forget” about the possibility of a severe crash, even though everyone “knew” that a lot of adjustable rate mortgages were about to reset and their was a lot of risk in the system. (Here’s a great take on the problem.)

In my previous post, I noted that using multiple assessments of risk, and modeling risk at the lowest level of granularity and aggregating up would give much better results than simply looking at aggregate data and having one expert pronounce their opinion. There are three other things you can do to mitigate risk that are especially important when the economy is weak.

First, don’t force customers into all-or-nothing choices. This type of bundling might be useful when customer don’t have a lot of alternatives, but when times are bad, it can lead to cancelation of entire orders or contracts, rather than a “graceful degradation.” Removing value-added services or reducing order volumes might be painful, but it’s easier to manage, financially and operationally, than suddenly losing big chunks of business.

This leads to another important process when modeling risk. Don’t just consider the financial impact of what-if scenarios– consider the operational implications. In other words, what might you have to do under certain scenarios? How might you adjust discount negotiation parameters if your company’s economic climate changes? How might you have to adjust production schedules? Thinking about this in advance avoids panicky decisions that can make the problem worse. For example, companies that develop strict discount negotiation policies during flush times find that their sales are slowing, and instead of adjusting their discount parameters, they throw them out, which keeps deals flowing in, but starves the company of vital profit. Companies that are not actively tracking sales, orders and their profit implications can suddenly find themselves in a bad position and they lay off the staff they might need to dig themselves out. Having a contingency plan and a dashboard would have allowed less drastic, more effective action, earlier.

Third, go back further. A lot of pricing decisions get made “because that’s how we always done it.” Go back to the folks who set up the pricing policies in the first place. What were they trying to accomplish? Does the policy still accomplish that goal? What risks were they concerned about? Have any sales reps who sold through the last big downturn? What did they do to survive that one (or not?)? Reaching back and asking the right questions will help avoid or mitigate recency bias and devise better scenarios and better contingency plans.

Comments are closed.