Less is More in Loyalty Programs

In approaching loyalty programs, too many companies think that they have to be all things to all customers. As a result they design programs that are either too expensive, or offer inadequate rewards.

The solution? Concentrate your attention on those segments of your customer base that will do you the most good, and do not waste significant resources on the others.

One cellular telephone company had a high rate of attrition, and declining revenue per subscriber. And it had a high proportion of low-end customers within its customer base. Its competitive environment was changing rapidly with price wars, new competitive technology and alternate distribution channels. It decided that it needed to:

*Reduce customer attrition.

*Increase customer lifetime value and customer loyalty.

*Get a greater share of its customer’s spending.

*Make a more effective allocation of their marketing budget.

The firm asked its service bureau to conduct two studies:

1. A segmentation and customer satisfaction study.

This showed that the best customers perceived the least value and were the least satisfied with the company’s service.

2. An attrition study.

This showed that:

*Defectors expressed low satisfaction months before the decision was made to defect.

*If a customer complained, the odds of defection quadrupled.

*If the customer was “very satisfied” with how a problem was handled they were quite unlikely to defect.

To identify potential defectors, the analysts:

*Developed 68 different statistical models exploring all aspects of the defection problem.

*Rated the models based on performance versus a control group.

*Selected a neural network for the final segmentation job which had the ability to handle a large volume of variables.

The key findings of the neural network model were:

*Two-thirds of all defections occurred within 15 months.

*Approximately 4 out of 10 defections were preventable.

*53% of preventable defections occurred before the seventh month of a subscriber’s contract period.

The stroke of genius was to use quadrant analysis to create customer segments, based on revenue and risk. This analysis really illustrated the segment where the company could achieve the most benefits from a loyalty program. It recommended that the firm create a program designed specifically for this critical group of customers.

Creating the Rewards Program

Based on the models, the company created a targeted rewards program focused on Group A: the high risk (high potential for attrition) with the highest current revenue. This group represented only 13% of the entire customer base. Other, less costly rewards programs were created for the other groups totaling 87% of the customers. The strategies were designed to:

*Identify the four key customer segments, adjusting them as new customers arrived, old ones left, and existing customers moved up or down.

*Allocate marketing investment based on revenue and profit for each segment.

*Provide different strategies and rewards to each segment within the loyalty program, and within that segment provide individual loyalty rewards based on a customer’s life stage, needs, and value.

*Provide super services and proactive communications to their best customers in Group A.

*Use the models plus customer daily behavior to detect defection problems and resolve them before the customer headed for the door.

The company established the recommended rewards programs, creating a control group for each of the four segments to measure the effectiveness. A year later, the firm found that:

*The programs generated a return on investment of $2.09 for every $1 invested.

*Attrition of those customers receiving the rewards communications in Group A was 1.27 points lower than those in a control group.

*Average revenue ($1,412) in the rewards test group was 5% higher than in the control group ($1,358).

*There was an increase of $19.6 million dollars in annual sales to those 13,881 customers who were retained by the loyalty program (compared to a control group).

This last result is the most impressive to me. The models were used to identify exactly how many customers did not defect as a result of the loyalty program.

Too often, a loyalty program is launched and conducted without ever knowing exactly how much good is being done—or not done. How can you prove how effectively any program is working? But setting up a control group that does not get the benefits. This is often the most difficult part of any rewards program.

This control group problem is illustrated by the experience of a master marketer a few years ago who explained:

“To prove that the money spent on the program was paying off, we had a control group. The company had never had one before. It is difficult to measure a program unless you have a control group.”

The marketer continued: “When we set up the control group, we made sure that no store executives were in it, no board members, no employees, but we did not have a smart enough database to tell us who was the next door neighbor and best friend of the president’s wife. She was in the control group.”

The result of all this?

“We got some angry calls from people who were in the control group. We didn’t tell them that they were in a control group, of course. We just told them that there had been “a terrible mistake”, and shifted them to the test group. These people were tagged as the “out of control group.” Others called to complain about being left out who were not qualified for the program. We explained that you have to spend so much to qualify. Most said OK, but for those who were adamant we made exceptions.”

In designing a rewards program like this one for the cellular phone company, the question often comes up; “How large should the rewards be to create the modifications in customer behavior that we are seeking?” One telco tried four rewards: a letter which thanked the customer for their business, a similar letter with 50 free minutes, and two other similar letters offering 100 and 200 free minutes. Which letter produced the desired behavior change at the least cost? The letter without the minutes produced only very slight improvements. The three letters offering free minutes all worked equally well.

Conclusion: Giving a valuable reward works better than no reward at all. But if you are giving a gift, it really does not matter how large the gift is. The company settled on fifty minutes as their most cost efficient reward.

From the above, it is clear that there is a lot that companies can do to reduce their attrition rate. The requirements: You have to have a marketing database that keeps track of customer behavior and the communications to the customer. The database has to hold the data that is necessary to create a predictive model. You have to experiment to find the most predictive models. Then you have to use your models to create segments that permit you to concentrate on where the problem is.

For your most critical segments, you have to experiment with rewards until you get the least costly, but most effective reward that produces the behavioral result you are trying to achieve. You measure your results using control groups. While this has been done by many companies, they represent only a small percentage of the companies that could and should be creating loyalty programs for their most critical segments.

Arthur Middleton Hughes is Vice President / Solutions Architect for KnowledgeBase Marketing. He is the author of Strategic Database Marketing 3rd Ed. (McGraw-Hill 2006).