One of the foundational ideas of business is the Pareto analysis that tries to identify the small portion of factors that are responsible for the majority of business results. You know the type—80 percent of profits come from 20 percent of the customer base and similar observations.
Vilfredo Pareto an Italian engineer and economist formulated a business classic that bears his name and we now simply refer to as the 80/20 rule. In fact the rule is malleable and subject to change based on all kinds of factors that are unique to a business. Some businesses see a 90/10 effect or even a 60/40 split but the point is that a small portion of all contributions to business success make the lion’s share of the contribution. But which ones are right for your business?
In a Harvard Business Review article, “AI Is Going to Change the 80/20 Rule,” analytics guru Michael Schrage points out three tips that can help any manager to do a better job leveraging big data and analytics. The one that most interests me is the idea that you can almost have too many Pareto analyses thus making the whole exercise full of noise and less predictive.
For instance what’s most important to keeping customer churn and attrition low? You could develop models, metrics, or KPIs that assess all kinds of things about your business from product quality to customer service to availability of online help—the list is endless. But only some of the analyses you come up with will be really important and it’s more likely that you’ll only arrive at a clear understanding if you track the interaction of multiple KPIs and their relationships with each other.
A couple of books ago, I wrote about the need for triangulation when using metrics and KPIs and Schrage seems to be on that path with his idea which is to perform Pareto analysis of your Pareto analyses. This, of course, sounds contradictory and it is but it also makes sense. Today a business can easily get to a point where it has not only too much data but too many analyses and when that happens going up a level of abstraction makes all the sense in the world—especially when you have the compute resources to perform the function automatically.
Note that this is different from using one or a small set of models at the department level. Those models are more tactical and can significantly help individuals to do their jobs better. A model that can tell a sales rep that this opportunity won’t happen but that one might is a real benefit because it saves the sales team from investing resources where they are not likely to be productive while focusing them on the better opportunities. But that’s not enough to run a business with and the purpose of the triangulated or Pareto-ed Pareto’s is to get a bigger picture that can include multiple departments and even input from customer communities to better understand how the company is performing over all.
This is a great example of two things, the power of modern intelligence systems to sift through a business’ voluminous data but also how technology advances open up new ways of thinking that were not available before. Not that long ago, a business leader might have had access to churn and attrition numbers and that leader might have even been able to see a month or two ahead to understand which customers were in danger.
Too often those danger signals were a green light to offer discounts on a future purchase. That sounds shrewd but what if the customer’s problem had nothing to do with pricing? What if the customer needed an upsell of services or additional products to make everything work. Schrage offers examples of this and I’ll leave it to you to look him up.
But today with this advanced approach, we’re becoming better able to understand the reasons why our primary indicators are flashing. With that we can be more confident about entering a customer interaction with relevant knowledge and we can drive our interactions toward conclusions that are more mutually beneficial.
(Cross-posted @ Beagle Research Group)