For many people, “predictive analytics” sounds like a solution looking for a problem. It feels more like something someone wants to sell to you, rather than something you need. You may have some knowledge of what it is or how it works, but until you can tie it to real problems it has no value to you.

However, I believe that there are some basic criteria anyone can use to qualify a business problem as a good candidate for predictive analytics.

Fundamentally, what predictive analytics does is gives the decision maker a new and better decision method to use, “prediction”, when making complex decisions.

This is the basic model for decisions I like to use:


I classify some decisions as “complex”, and the rest as “routine” (i.e. not complex). When the decision to be made is complex, the decision maker can benefit greatly from having a new and better decision method to use when compared to guessing or going with gut feel.

So if you look for and find a business problem for which you can answer “Yes” to these 6 questions, you have found a complex decision that is a good candidate for predictive analytics:

  1. Does this decision have many (input) options to choose from?
  2. Does this decision require multiple (output) choices to be made?
  3. Is the total cost of action is not trivial?
  4. Is it critical that the dollar value of results exceed the total cost of action?
  5. Do you have little or no information to support this decision?
  6. Do you already have, or can you collect, relevant data to support this decision?


  • Customer churn – which of my current customers are more likely to leave in the next 12 months? If I knew that in advance, I could intervene with my best customers and try to prevent it.
  • Employee turnover – which of my current employees are more likely to resign in the next 12 months? If I knew that in advance, I could intervene where I had a high value employee.
  • Collections – which of my current customers with overdue bills are more likely to pay up with a nudge? If I knew that in advance, I could have my A/R team follow up with them first to reduce my DSO more quickly.
  • Predictive maintenance – which of my many production machines is more likely to fail and create downtime? If I knew that in that advance, I could focus my resources on those machines first and improve production uptime with lower cost of maintenance.


Guessing or going with your gut may work well enough when making routine decisions, but those decision methods can be expensive, unproductive, or both when you need to make a complex decision. In those cases, you need a new and better decision method…prediction.

If you’d like help finding business problems with complex decisions in your business, contact me at and let’s talk about what’s possible with prediction.