Customer trade-offs

The mail order company run by Roy Cardiff keeps track of each customer's transactions. Recently, he decided to cut costs by reducing the distribution of catalogues to customers who are less eager to buy. Cardiff's customers can be divided into three categories: customers who have placed multiple orders in the past year but with a small amount; customers who have only placed one large order; customers who have placed sporadically for a long time.

Which type of customer should he remove from the mailing list? According to several marketing experts who have studied the issue, the solution to the problem is not that simple, despite new and increasingly sophisticated ways to measure "customer lifetime value (CLV)." The so-called customer lifetime value refers to the present value of future revenue that may be brought by a single customer.

Using Data to Discover Customer Lifetime Value

“CLV is hot right now,” notes Wharton professor Xavier Dreze. CLV is not a new thing, but with the rapid development of Internet technology "enabling companies to have direct contact with customers at a lower cost", the concept of CLV is also becoming more popular. CLV "sees the customer as the source from which the company strives to get as much value as possible," Drez said.

Yet many companies are finding that CLV, which is part of customer relationship management (CRM), remains elusive. First, there is a lack of certainty in measurement; second, it is difficult to use in practice.

“The only data managers really have confidence in is the profitability of their customers today,” said Wharton professor George Day. “The fundamental problem is that you have data now, what How to use it There is a good chance that other customers will be dissatisfied."

Furthermore, it is difficult to predict how long a customer will remain in a transactional relationship with the company, or how "growth" he will be. "In an analysis of growth, it's difficult to know exactly how much profit a customer is likely to bring," Day points out.

In industries with higher costs of acquiring and retaining customers, such as financial services, air transport In the industry and hotel service industry, CLV can play a greater role. “When transactions are distributed asymmetrically—that is, when most of the turnover is generated by a small number of customers, as is the case in the hospitality industry; or when businesses are able to influence consumer behavior through incentives or other incentives, CLV also more effective,” says Wharton marketing professor David Bell. For example, airlines can attract passengers by adjusting passengers to first class, and the passengers get a lot of benefits, while the increased costs for the airlines are minimal.

Bell also noted that certain companies could benefit greatly from collecting CLV-related information. For example, the customer stay information collected by a hotel can help it identify the best customers and cross-sell other products to them. Marketers of hotels can get feedback from these customers. Based on feedback, hotels can make rational decisions about how to allocate marketing resources more efficiently. Assuming the information shows that the majority of customers are from the largest cities and are in their fifties, the hotel can use that information to accurately expand its business.

Bell sees Harrah's gaming business as a good example of a successful use of CLV. Based on information obtained from customer loyalty programs, Harrah's can find out "who will enter the company's gaming halls, which game will they play first when they come in, how long they will spend at different gaming tables, etc. etc. With this information, the company can make appropriate adjustments to the variety and settings of the games.”

Other industries that may benefit from CLV data include: Healthcare, Credit Cards, Direct Marketing, and the Internet The email marketing industry, in part, because they are all in direct contact with customers, it is easy to record customer information. Drez cites the pharmaceutical industry as an example: Salespeople can use relevant information to decide how often they visit doctors for drug promotion.

Day says, "Usually, CLV is more usable as long as you have data like customer profiles and transaction information. But if you don't have a direct contact with the customer, but through something like a retailer that adds value, If you sell through channels like CLV, then using CLV is not so simple."

Customers are changing, and subjective judgments must be added

After collecting better customer purchase information, it is enough to judge the customer's lifetime value, How to use this data? The researchers advise: "Cautiousness is advisable."

"Everyone is different," Bell said. "Analyzing individual customer behavior is harder, and predicting behavior patterns for a market segment is easier. Let's say that the average business traveler spends several nights at a Hilton hotel. But it's hard to say exactly how many nights Mr. So-and-So will be staying at the Hilton."

Bell also pointed out that one of the difficulties with applying the CLV method is that the predictive model is very sensitive to assumptions. For example, the model often assumes how long the customer will maintain a business relationship with the company, whether the customer is active in spending, and how much the customer will spend. However, some of these assumptions are inaccurate. "Just because I spent $100 last year doesn't mean I'm going to buy a $100 order this year," Bell said. "It's also important to be clear: a customer is less active because he temporarily stopped using the product. , or did he switch to a competitor's equivalent?"

The problem with previous approaches to assessing the value of the Internet was that many companies made inaccurate assumptions about what a customer was worth and what the cost of acquiring a customer was. How much, how long the customer will maintain a business relationship with the company, etc. "And the results of the value calculation are very sensitive to these important assumptions." Bell said, "Any mistake can have a serious impact, that is, an assumption that does not match the actual conditions may make the calculation results vary widely. But A lot of companies these days base their service levels on the lifetime value of the customer. If I'm a regular customer, I'll be dismissed; if I add two stars, they'll serve me with respect. They do this because they assume that the customer is Static: You put customers in a certain group and they just stay there. But the reality is: If you had a better attitude at the beginning, maybe I would have been a big customer."

In addition, when companies assess the value of their customers, it is customary to make inferences based on the customer's historical transactional relationship with the company. "The information is incomplete. You don't know the customer's transactional relationship with other businesses. Maybe he spends $100 a year with you, but at the same time buys $500 of your competitor's product." Bell said, he Refers to the "share of wallet" -- that is, how much customers spend on your products and how much they spend on your competitors' products. Here's the problem: You can't just use the data of a customer's transactions with one of your companies to value him.

Any predictive model employed by a company provides only one aspect of the decision-making process. "It's up to intuition and managerial judgment," Bell added.

Day gave an example: A large component manufacturer found that a certain customer was not profitable. “What do you do in this situation? The customer may not be profitable, but in this market, it may be 15% of your turnover. The cost of abandoning these customers is huge. Customer lifetime value has to give way. Predicting the future value of a customer is tricky: How do you know what a customer will do in the future?” Day said the company’s biggest risk is “inadvertently abandoning a customer that will be profitable for the company in the long run. ".

Feder points out that some CLV measurement models ignore the "inherent randomness" of individual customers. "These models only look at past customer behavior, assuming that each customer will deliver a relatively fixed benefit at a particular stage. But information from past transactions is not the best or only factor for predicting the future."

Marketing Alignment, Success or Failure

Marketing strategies such as cross-selling and up-selling have been around for many years, and more and more businesses are now actively using them to increase customer lifetime value. However, success or not, it is difficult to generalize.

In the case of cross-selling, the company that sold you your water skis will surely also sell you goggles. For marketers, the appeal of doing so is obvious. "It's much easier to sell to old customers," says Drez. "It's about maximizing the value of existing customer relationships." But Feder is somewhat skeptical of this strategy: "If the customer is largely If he buys at will, then it is hard to say that there is a necessary connection between his random purchase of product A and his random purchase of product B."

Upselling is also problematic. Take Amazon as an example. After a customer buys a book of several dollars, Amazon will waive the delivery fee; or after the customer buys the first book, it will provide a discount on the second book. "In the case of Amazon, there are customers who might not care about the discount on the second book and pay in full," Feder said. "Some companies put too much effort into upselling. It's hard to say what real effect these efforts have had. The increase in sales does not suggest that the improved profitability can be attributed to marketing efforts."

with Cross A similar strategy for sales is multi-channel marketing. "In the past, most companies had only one point of contact with their customers," Feder said. "But now there are many retail channels, as well as Internet, direct mail, and call centers. That requires allocation of resources. If Should one customer use the internet and the other use a call center, should it be handled differently? Obviously, you want to encourage more customers to buy through the network, because the maintenance cost of the network is much lower than setting up a call center. The question is, should What kind of customers are encouraged? What are their behavioral characteristics? Do you risk angering call center customers to encourage them to order online? Or do you go all out to mobilize customers who are less enthusiastic about the call center approach to go online, even if the business is not profitable. And increase?"

There are many ways, don't stick to quantitative data

Feder said the problem boils down to this: "Some sales strategies are good, and some sales strategies are not very effective, but in general, it is difficult to get from marketing Identify the portion of the return on investment that can be attributed to the ongoing implementation of CLV management measures. Companies have implemented a variety of sales strategies to attract customers, inadvertently reducing the validity of CLV data, making it more difficult to This will determine which customers to take and which to leave in the future.”

In a new research report, Feder and Bruce Hardie, Chun-Yao Huang and Ka Lok Lee) analyzes how marketers using customer information as a starting point assessed the value of different customer groups based on past behavioral patterns prior to the widespread adoption of CLV by managers. "The most common approach is to categorize customer value prospects in terms of RFM [referring to recency, frequency, and monetary value of transactions]," says Feder.

RFM is actually derived from the direct, most widely used concept of CLV. Marketing industry. Feder and co-authors wanted to see how the relatively simple measure of RFM correlated with the more complex CLV value, which is likely to be the "prime indicator" for predicting consumers' future buying behavior. "If one customer purchased many items a long time ago, and another purchased one item recently, which customer is more valuable from a CLV perspective and is more worthwhile to keep?" Feder added. The case at the beginning of this article, "How should the importance of transaction recency and frequency be balanced?"

In their report, Feder and co-authors point out that, in fact, simple statistics such as transaction recency and frequency can Provides a valid estimate of future lifetime value. He said: “With limited and refined transaction information, when used properly, it is possible to produce accurate CLV predictions as well as full detailed transaction information. The challenge for marketers is to determine which inductive statistics to use, and How to use correctly judge. In other words, despite the availability of modeling tools that use transactional information to predict customers' future buying behavior, managers still rely heavily on subjective judgement when identifying the best and worst customers to drive future sales growth.

Feder found that managers were not used to sticking to inductive data like recency, frequency, and monetary value. The way they treat this data depends largely on the nature of the task at hand (in the previous example, deciding which customer names to add and remove from a mailing list), and the form in which consumer purchase information is delivered to them. "In order to avoid managers who do not accept this information and fall into 'black box operations', it is important to understand how external factors will affect their decisions." Feder said, "We are working hard to build high-tech predictive models and better Learn how managers find the balance between psychological factors in management decisions."

Dreze and Andre Bonfre propose a "new approach to assessing customer value." “Traditional CLV measures the net present value of all revenue a customer generates. Part of the assumption that marketers make when predicting lifetime value is that at some point customers will churn and choose other companies’ products.” Drez said.

But in making this assumption, "you underestimate the value of the database. If you optimize your marketing efforts in this way, you will make the wrong decision. Because, while you lose some customers every year, you also New customer acquisition. New customer acquisition must be taken into account when evaluating the value of the database,” added Drez. "It's about maximizing database value -- not customer value."

And Wharton operations and information management professor Noah Gans analyzes CLV from an optimization perspective Question: If resources are limited, which type of customer should be centered?

Ganz developed a theoretical model to study how the overall level of service quality affects the average length of time that customers maintain a transactional relationship. "If you increase the average quality of service, customers will stay in a transactional relationship with you significantly longer," he said. But there are other questions to consider: What are your competitors doing? What is the cost for customers to switch to other providers? How will the evolution of technology affect purchase transactions?

Sometimes, a company makes inferences about what type of customers it's dealing with. “The company then decides to provide that customer with a certain level of quality of service. Like in a call center, that customer may be served first. This is the operational control the company uses to manage the quality of service the customer receives and the cost of serving that customer. measures.” Ganz acknowledged that competing suppliers would naturally agree on a service-level “standard.” "In the real world, slogans like 'world-class service level' that you hear a lot reflect this standard," he said. "Take the call center business as an example. 80 percent of calls have to be around 20 seconds. This is a generally accepted standard.”

On the issue of maximizing CLV, Ganz believes that it is more effective for companies to record the purchase information of each customer and On this basis, customers are classified. “Then, based on what you know about the characteristics of that customer group, decide what level of service to offer, whether to cross-sell, up-sell, or any other marketing strategy you see fit. When a customer comes to buy, you don't really know which group they belong to. So, the optimal decision must take into account your uncertainty about the customer's response."

This article is adapted with permission from Knowledge@Wharton (http:/ /knowledge.wharton.upenn.edu) July 30, 2003 Article "Which Customers Are Worth Keeping and Which Ones Aren't? Managerial Uses of CLV," The Wharton School of the University of Pennsylvania Copyright 2003. Translated by Liu Songjie.

Free Alerts on newly added products!