Corporate Information Factory (CIF) Resources by Bill Inmon, Inmon Data Systems

> home > view content

Managing Customer Loyalty: Getting Smart

By W. H. Inmon

After the novelty of sending messages is over, it isn’t the way the message is sent and delivered; it is the business content and value in the message that separates success from failure.

No truer words have been spoken and they can be directly applied to the Internet environment. The business that is done over the Internet quickly shifts emphasis from the novelty of operating in a new technological media to an emphasis on dollars and sense and profitability of the business. As such, the Web-based Internet eBusiness environment is a business accelerator. The Web environment will accelerate your business success or it will accelerate your business failure. In the long term, what is important is the business model behind the Web environment. If the business model behind the Web environment is sound, then the Web will make the business model even more sound. If the business model behind the Web environment is unsound, then the Web environment will make the business model even more unsound.

If the technicians of the world and the venture capitalists had paid more attention to the business model and less attention to the “gee whiz” aspect of Internet technology, then it is unlikely that there would have been the economic bubble burst of recent vintage. Instead, the technicians and the venture capitalists got so wrapped up in the technology of the Internet that no one asked the hard question – does all of this make business sense? From March 2000 to March 2001 it is estimated that 90% of the venture-backed Internet companies have disappeared. Undoubtedly, most of them disappeared because of this one fundamental flaw.

The Internet is good at managing and delivering messages. The Internet is a wonderful delivery mechanism. But the Internet doesn’t do anything magic. If all that flows across the Internet is a bunch of meaningless messages, then the Internet simply isn’t worth very much. It is up to the business to put the beef – the smarts - into the messages that flow across the Internet. In short, the Internet is good at delivery, but it is up to business to put the intelligence in those messages that pass up and down the Internet.

So how does a business arrive at the point of making smart messages? If in the long haul it is the message that counts, not the way the message is delivered, then what makes for a smart message? One important aspect of a smart message is in knowing who you are conversing with. If you don’t know anything more about the person than merely their name, then you are at a severe disadvantage. Wouldn’t it be nice to know a lot about who you are conversing with? In particular, wouldn’t it be nice to know whether the person you are talking with is a person that is happy with the business relationship or conversely, someone who has generally been unhappy? If you knew the happiness level of the person you are conversing with, then you could have a very different conversation. Happy people get to be treated one way, unhappy people treated another way. And you need to know this before you even begin the conversation. Knowing who your customers are, what they are worth to you, and how happy they are with your business relationship is a good starting point for the creation of smart messages.

What information then does a business need in order to get to know its customers? How do you tell a good customer from a bad one? A loyal customer from one who is likely to be disloyal? A profitable customer from an unprofitable customer?

One of the most important of these issues of “smartness” is that of customer loyalty. It is well known that it is much less expensive to sell to an existing customer than it is to sell to a new customer. Therefore, when we are sending messages across the Internet, the more we know about existing customers, the better chance we have of selling else to them and further extending their loyalty. Protecting customer loyalty is one of the most important things any corporation can do.

The story of determining a customers loyalty begins in the data behind the Web, the data warehouse. The data warehouse is a collection of detailed, integrated, historical data. Typically a data warehouse will contain a lot of information about the past transactions the corporation has had with a customer. Among other things the data warehouse contains much information about customers. Built properly, a data warehouse tells such things as:

When the customer first signed up for service
How often the customer partakes in a service
What services a customer uses
How much money a customer spends
Where a customer uses a service
When a customer uses a service, and so on.

In addition, in a fully developed data warehouse there is likely to be much other information about a customer such as:

Age, gender, personal information
Economic information
Home location
Family
Occupation, and so forth

The data warehouse will carry this integrated data for a lengthy period of time. Depending on the environment, detailed customer activity may go back up to five or ten years. It is this integrated, detailed historical data that is worth its weight in gold when determining the loyalty and disloyalty of customers.

The first step in creating a loyalty analysis environment – sometimes called doing churn analysis - is to find out which customers have been loyal and which have not. Since the first date of service and the last date of service (if the customer has taken his/her business elsewhere) are a natural part of the data warehouse, it is easy enough to find the top ten percent of customers who have been with the company the longest time and the bottom ten per cent of former customers who have been with the company the shortest amount of time. This is done by a simple analysis of the service dates in the data warehouse.

These customers are selected and put aside.

Next the records for these customers are gathered and analyzed in order to determine what common characteristics the customers have. Some of the analysis might look like:

Do loyal customers live in a particular area?
Are disloyal customers mostly women?
Do loyal customers sign up for first service on weekends?
Do disloyal customers use services infrequently?
What age are loyal customers?
Do disloyal customers own their own home? And so forth.

The characteristics for customers – both for loyalty and for disloyalty – are studied closely. The integrated, historical records are looked at many ways. Some perspectives will be fruitful and others won’t. For some categories there will be the same, or nearly the same, set of characteristics for both loyal and disloyal customers. But for other categories of characteristics there will be a sharp difference between loyal and disloyal customers.

Once the categories are collected and analyzed for both loyalty and disloyalty, the next step is to create a profile. The profile is based on the correlation of different characteristics that are associated with loyalty and with disloyalty. For example, a profile might be created that looks like:

Loyal customer

Male
From 35 to 45
Salaried
Owns home.

Disloyal customer

Female
From 18 to 26, from 45 to 60
Unemployed
Rents

The profile for a loyal customer and a disloyal customer has now been established. What can a company do with such information?

One obvious usage of the profile information is to go back into the existing customer file and determine how existing records match up to the profiles. For each existing record a data element called – CLASSIFICATION – is created. Then the profile is passed against each record and an assertion is made about the existing customer based on how the customers records compare to the profile. The matching process is hardly perfect. Very few customers records will be a perfect match to the profile. For example, suppose the match finds a female that is salaried, owns her own home, and is between the ages of 35 and 45. The criteria fit for a loyal customer except that the customer is a female not a male. The program that does the matching and qualification must be able to handle less than perfect matches.

Each record in the data base is assigned a value in the CLASSIFICATION data element of a “d” for disloyal, “l” for loyal, or “-“ for undecided. Now the company can tell at a glance exactly how many of its customers are at risk and how many are not. In addition, the company can do many things to keep the services of disloyal customers. The company can do such things as:

new programs with special and unique services
Offer to cut the price of services
Offer individual attention, and so forth.

Since the company knows exactly who is at risk then the company can do what ever is necessary in order to keep the marginally loyal customers in the fold.

But there is another way the company can use the profile information. It is one thing to classify existing customers based on what you know about the customer. It is another thing to predict the behavior of a customer when he/she walks through the door for the first time. When the company has the profiles of what a loyal customer and a disloyal customer look like in hand, the company can make a very educated guess as to the long term viability of a new customer when he/she first signs up for service by comparing the new customer’s traits against the loyalty profiles upon entry into the system. It is kind of like reading a crystal ball. The approach is admittedly imperfect. Some customers will simply be pegged incorrectly. But on the average, the odds are good that the company can predict who will and will not be loyal upon entry of that person into the system.

Based on the prediction, the company can take actions to make the marginal and likely disloyal customers into loyal customers. And by making disloyal customers loyal, the market share of the company goes sky high. Other companies have a hard time trying to take a customer away from the company that understands who is likely to be loyal and who is unlikely to be loyal.

Once the profiles are created, the corporation can then use that information creatively in a thousand ways. The limitations are only in the minds of management and users.

It is noteworthy that the profile activity needs to be periodically revisited. Over time the profile for loyalty and for disloyalty changes. The company needs to be sensitive to those changes.

Once a customer gets to be “locked in”, the churn of the company is reduced. Customers are likely to be entering the system but not leaving the system with any degree of regularity. And once market share is established and is defended, then the business of the company is put on a firm footing.

And how does the knowledge of what qualifies a loyal customer or a disloyal customer have to do with messages sent across the Internet? Plenty. The messages sent across the Internet can vary. Loyal customers can be sent one kind of message and potentially disloyal customers can be sent a different kind of message. Each type of message is tailored to the projected proclivities of each customer. Now that’s getting smart.

The Internet is a powerful medium, but to make it effective, the messages sent across must be “smart”. There are many dimensions to smart, but perhaps the most important dimension is to know who you are conversing with and the proclivities of that person. Once you start to be smart about the person you are having a conversation with, then all sorts of possibilities enter the picture.