Measuring Distributor / Agent Loyalty in Service Businesses

Jim answers questions from fellow Drillers
(More questions with answers here, Work Overview here, Index of concepts here)

Topic Overview

Hi again folks, Jim Novo here.

Today we have a fellow Driller looking to compare the “loyalty” of sales / distribution agents for insurance products and use this information to manage business with the agents more effectively. In this case, knowledge of the business is exceedingly important because segmentation of business lines across regions will dramatically improve predictions.

Let’s do some Drillin’!

Q: Hi Jim,

I happened upon your site and found the information there very valuable – so much so that I ordered your book (customer is referring to Drilling Down).

A: Well, thank you very much for that!

Q: I’m a marketing manager with an insurance company that distributes its life, auto, home, and business insurance products through independent insurance agents.  These agents represent our company as well as others.

I’m interested in techniques for measuring agent loyalty – which I think would be demonstrated by the agents choosing to place business with our company instead of another company they represent for policies.

A: I’m not sure in this case anything is too terribly different from the scenarios used in the book. Essentially, agents or consumers demonstrate loyalty though their actions, and if you can track their actions, you can spot increasing or decreasing loyalty.  Your business is more complex in many ways than retail, but to the consumer (in your case agent), there are always choices to be made between alternatives, and changes in the purchase patterns agents or consumers generate often precede customer defection.

In a very simple case, let’s say the average agent writes a policy every week with you.  Some will write more, some less.  But what you are interested in for estimating loyalty (increasing or decreasing?) is not the rate at which they write policies, but any change in rate.  If you have an agent writing 3 policies a week and they drop to 1 a week, this is a significant change in behavior, and this behavior should be flagged and investigated as a prelude to agent defection. 

If this agent is a “best agent,” then the need to find out if there is a problem is even more urgent.  The more policies the agent writes, the more imperative it is to find out if something is wrong.

In the book, the Recency (how many days since last policy was written) and Frequency (how many policies have been written in total) of writing policies is used to rank all agents against each other for “likelihood to keep writing policies.”  Any changes in this likelihood show up as a change in rank – called RF Score, or Recency-Frequency Score – and will alert you to high value agents who may be defecting and dropping your lines.  There are several different versions of this approach; you can read about one of them in some detail right here on the site.

Depending on the data you have access to, another approach is to use Latency, in which simpler average behavior patterns rather than agent scoring are used.  The example here would be the average agent writes one policy a week, and those who slide below this rate are likely future defectors.  You can run these Latency numbers by line by area of the country for example, because the average Latency of writing a Life policy in New England may be different than for a home policy in the Southeast or auto policy in California.  For more information, see this tutorial.

If I am way off base in understanding how your business works, please let me know.
(Jim’s note: she didn’t contact me, so I guess I was on target)

I hope I answered your question!


Get the book at

Find Out Specifically What is in the Book

Learn Customer Marketing Concepts and Metrics (site article list)

Download the first 9 chapters of the Drilling Down book: PDF 

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.