Developing the Personalization-Centric Enterprise Through Collaborative Filtering and Rules-Based Technologies

PDF File: 

CRM1_wp_larsen.pdf

NetPerceptions

Businesses that choose a personalization-centric mindset as integral to their CRM initiatives have a variety of technologies at their disposal. Two of the most widely deployed are collaborative filtering and rules-based systems.

Executive Summary
The following presentation offers CEOs, COOs, CTOs and other top executives a new approach to business that was not possible as recently as five years ago but has become mission-critical in the highly automated, intensely computerized business arena of the new economy. It is the concept and implementation strategy of the personalization-centric enterprise.

A personalization-centric enterprise addresses Customer Relationship Management (CRM) as a corporate philosophy by integrating personalization applications throughout the business. Details of benefits to both customers and companies appear in this text. Included, too, is an explanation of how collaborative filtering and rules technologies make very precise personalization the most profitable approach for any enterprise. Corporate visionaries worldwide are acknowledging the potential – and the results – of the ideas presented here by Net Perceptions, the award-winning leader¹ in personalization technologies.

Definitions

• Personalization — the act of customizing the presentation, display or offering of a product or service to a particular person.
• Personalization-Centric — an enterprise, process or ideology in which personalized products and services are integrated and implemented throughout the organization, including all points of sale, other points of customer contact and back-end activities and departments, such as inventory, shipping, production and finance.
• Collaborative filtering — the computerized technology that matches the preferences of a single customer with a database of other customer preferences in order to deliver accurate recommendations of products and services to that individual consumer.
• Rules-based technology — an earlier form of personalization technology that presented products or services to groups of individuals according to standards that were established by marketers in advance to be associated with those groups.

Introduction: New Value in Personalization
The explosion of electronic commerce has brought personalized marketing to the forefront of e-commerce Web sites. It has been identified as a critical factor in Internet business because it is one very significant way that users can differentiate between Web commerce sites.

Internet shopping has radically changed the consumer purchasing experience and companies must adapt or die. It turns out that even though consumers want the best price, quality and convenience, those elements of the purchase experience are rapidly achieving a level of equilibrium through the Internet. So, businesses are asking: Why does a consumer buy from one Web site instead of another?

Industry pundits Joseph Pine and James Gilmore tell us that the shopping experience, the pleasant and successful interaction between human and machine, is more important than what the consumer is actually seeking to purchase.²

Personalization makes the difference. Personalization allows a business to match products or services with the particular individuals who are most likely to value and want to buy the products or benefit from the services. This gives each user a unique experience (even though it may not be obviously unique to the customer).

CRM Beyond the Web Site
Heretofore, personalization has been recognized as a way of letting customers know that you know them. There are four general categories for this type of knowledge: 1) name recognition, 2) check box, 3) rules-based and 4) preferences-based personalization systems. These categories still exist, but the concept of personalization has expanded in scope to embrace the much broader notion of CRM. In addition, contrary to the popular belief that the above systems are competitive and exclusionary, they are, in fact, complementary methods that can be very effectively combined.

CRM takes personalization from a Web site technology to a corporate philosophy. Today personalization is an ideology that is being adopted by visionaries. Any company that already has such a forward-thinking person as its CEO, president or board member is a step ahead of the pack. But personalization-centric as a way of doing business is a new concept. The idea of linking all customer touch points to a single database is just now being introduced as a highly successful business practice.

Don Peppers and Martha Rogers published this idea to the business world in The One to One Fieldbook: The Complete Toolkit for Implementing a 1-to-1 Marketing Program.³ They suggested an application that was already in the development pipeline at Net Perceptions, namely that the most efficient and profitable way to run a business is for the production, inventory and delivery parts of an organization to be directly responsive to feedback from the sales and customer service departments, based on real-life information provided by customers. By listening to each customer and incorporating that data into the enterprise system, true one-to-one marketing emerges.

When a business is personalization-centric, the management of customer relations occurs at all levels and in all departments of the business. Not only does the Web site use personalization technology, but the entire enterprise uses personalization to make constant, up-to-the-minute decisions for each transaction. The Internet has inspired new technologies that vastly expand the possibilities for personalized contact with every customer.

Peppers and Rogers call it "functional integration." We refer to it as "personalization-centric."

Evolution of Personalization
Early personalization methods worked on a set of profiles, categorizations, segmentations and rules of average behavior. Those rules of market segmentation functioned by identifying common denominators such as zip codes, income levels, education levels, marital status and gender and then placing all the names in the database into a sorter mechanism that let the names filter down into the common denominators. Products were targeted at specific market segments based on the likelihood that products could be sold to people in those segments.

The new personalization technologies work in the opposite direction. These systems take the individuals in the database and analyze each one to find the similarities between them, without predefined categories of similarities. This change in approach has proven, among other things, that the previous segments thought to define groups of people actually don't define them well at all. What really works in defining types of consumers is identifying their preferences: tastes in music, colors and models of cars, restaurants, books, vacation activities, and behaviors which can be found in purchase histories. The affinities in personal preferences overlap all previous borders artificially set by income, gender, race, geographic location, etc. A business using preferences-based personalization can offer products to specific individuals based on a likelihood of those individuals actually buying those specific items.

A Big Difference
What may appear to be a fine line of distinction between "likelihood to sell" and "likelihood to buy" is actually a great chasm. Past systems developed profiles, discovered or created a limited number of categories, focused on mainstream demand, sought the average consumer. Present systems acknowledge that the average consumer doesn't exist; only individual consumers exist, and they have some preferences in common with other individuals. It is a vastly different approach to say that two individuals with similar preferences (enjoy the same movies, enjoy biking and like Thai food) are likely to buy the same item than to say that two individuals in the same geographic category (income level, zip code or gender) or same psychographic category (new-ager, techie, early adopter) can be sold the same item.

In fact, it is a revolutionary change. The new personalization technology is individually pertinent, it is technologically driven, and it delivers a unique experience. How does this relate to your business, both in the Internet space and out of it?

Learn from Every Contact
A personalization-centric business makes sure that every contact with every customer is a learning experience. Every purchase, every sale, every return, every request for information, every indication of preference – every contact, and as much ancillary information as possible relating to individual customer preferences – is assimilated into the personalization engine, the heart, as it were, that drives all other actions. Data from all these customer interactions is then used across the enterprise to determine additional customer acquisition and retention activities, manage products for the customer base, manage credit risks, and create buying opportunities for each customer via each customer touch point. Customer service representatives who handle sales and returns record their customer interactions to help improve their own capabilities for future activities, while the same information is also helpful to management, personnel, shipping and scheduling.

All parts of an organization are dependent on all other parts. In most organizations, the differing segments are held together by the paperwork shuffle and word-of-mouth communications. These communication systems are unique to each company and strongly influence efficient, cost-effective operations and pleasant working environments. With personalization technology connecting all departments, these elements of business are seamlessly improved to the benefit of management, stockholders, employees and, most importantly, customers.

Personalization from the Customer Viewpoint
Why become a personalization-centric enterprise? Consider the customer. The benefits that accrue to the customer are among the most important factors ever identified as leading to customer satisfaction:

1. An expedited road to the products and services that the customer desires
2. Time saved in locating the right product or service
3. Learning about new, relevant products or services
4. Personal recognition and attention to detail that make subsequent purchases or product returns pleasant, fast and easy
5. Personal recognition that makes promotional material relevant instead of irritating.

How does personalization-centric technology deliver these benefits?

1. Matching customer needs and desires with available products and services is the magic of preferences-based personalization applications like Net Perceptions for E-Commerce. The scope of the technology's capabilities is so immense that literally millions of users (for example, all of the registered card holders of a department store chain or library card holders) and their individual preferences and purchase histories can be instantaneously cross-referenced with tens of thousands of products. With these billions of bits of data tumbling around in the hopper, a preference-based personalization system can extract recommendations uniquely pertinent to each and every individual in the database, while the interaction is in progress.



2. Time is saved both by helping the customer locate the best products in the offering mix as well as by presenting those items in real time, while the customer is right there making the transaction. And real time product recommendations can occur in a variety of places, including the Web site, a customer service desk, a call center, a kiosk in a store, and others.



3. Learning about new products is a significant benefit to customers that formerly was exclusive to the in-store browsing/shopping experience. For example, a customer looking for something new in music that would suit his taste has a hard time finding the right item without a recommendation from a trustworthy source, one that knows him and his tastes. A recommendation engine is the most effective way to present that buyer with new material that will not only fill his need but also delight him with a new avenue to explore further.



4. Personal recognition during future transactions gives the customer a feeling of confidence and comfort: "these people know me." A customer returning a product enjoys a receptive customer service interaction based on the customer service representative knowing how best to relate to this individual customer. A customer returning to buy another product likes to have the business remember previous purchases because that speeds up the decision-making process of the next transaction and ensures the right choice for that customer.



5. Personal recognition in promotional activities improves the customers' opinion of the company by directing only pertinent material to them and not loading the mailbox (either electronic or curbside) with "useless junk." Highly relevant, individualized recommendations are viewed not as marketing-hype, but as a valuable customer service. For example, "Ryko Records has a new 3-Disc retrospective on the music of Bill Evans," is viewed as valuable news to a jazz fan with a special liking for the piano playing of Bill Evans.

Continuous System Learning
Repeat success on these five factors depends on another aspect of today's personalization technology, the ability to learn. The system gets smarter with each individual customer interaction, allowing the enterprise to more precisely meet the customer's needs and desires.

With personalization integrated throughout the company, knowledge of a customer's preferences can be at the fingertips of any customer service representative or salesperson in any store location or call center, as well as on the Web site. This integrated approach makes it possible for the success factors to be met at any of the touch points where the customer interacts with your company.

Personalization from the Enterprise Viewpoint
Why become a personalization-centric enterprise? Consider the effect on the company. The benefits that accrue to the company are among the most important factors ever identified as meeting commercial business objectives:

1. Repeat business
2. Reduced costs in serving each customer
3. Increased cross-sell and up-sell success
4. Successful promotions

Personalization helps to satisfy these business challenges:

1. Repeat business. The more completely you satisfy your customers' needs, the more likely they are to become loyal customers. Preference-based personalization increases the chances of meeting a specific customer's needs by ensuring that the products presented to that individual are the ones from your inventory that are most likely to meet the customer's conscious and/or subconscious needs or desires. Additionally, the clearer it is to the customer that he is known at all of your corporate touch points (Web site, store, ticket counter or call center), the more likely he is to visit multiple touch points. The more likely a customer is to visit multiple touch points, the more likely the customer is to be loyal and profitable.



2. Reduced costs per customer. With fierce price/cost competition and thin margins dictated by the extreme efficiencies of e-commerce, budget allocations for acquiring and keeping customers have to be effectively spent. Thin profit margins also increase the necessity of cross-sell and up-sell opportunities, a very difficult objective to achieve on the Internet without a successful recommendation engine. Automation is the only way to address each customer as an individual when face-to-face contact isn't possible. Even in a face-to-face situation, the salesperson who has access to the customer's purchase history is better able to serve the customer on a personal basis. Using personalization to know your customers, learn more about them and increase your service to them addresses this business challenge aggressively and successfully. Companies can now provide a level of service never before dreamed of, and at an incredibly low cost, because it leverages unexploited assets (other customers' preferences) so effectively. The capabilities of this technology far surpass all previous one-to-one sales situations because the sales person – any company representative or a non-human touch point such as a Web site – can tap into so much useful information to help the customer get exactly what he wants. And this is economically feasible because it's automated.



3. Optimize cross-sell and up-sell opportunities. This is especially valuable to firms with large inventories selling over the Internet. When a business has thousands, or hundreds of thousands, of products to sell, the Internet can actually be an impediment to the customer who's browsing for an item. Parsing an inventory list into subjects or categories helps but still requires the user to spend time reviewing lists, waiting for downloads and perhaps giving up before finding the desired item. A good personalization system, however, allows the user to browse the complete inventory list while also recommending specific items that are likely to appeal to the user. Quick, accurate recommendations demonstrate to the customer that the site is responsive and fulfilling. Accurate recommendations have proven to increase sales by inspiring the customer to buy not only the recommended item but also ancillary items that relate to the recommended one.



4. Successful promotions. The Internet provides businesses with a number of "mixed blessings" and one of them is the increased scope of your potential customer base. However, with more customers and more competition and low profit margins, finding the correct time, medium and product selection with which to approach each individual customer can't be done without an automated personalized process. Personalization in the back office – that is, steeped in all other parts of the enterprise – ensures that the personalization that works "up front" is more than skin deep. Personalization technology applied to advertising, direct mail promotions, customer service, inventory, even shipping, improves the applicability of any item to the promotion recipient.

Making It Happen
Why become a personalization-centric enterprise? Consider the effect on the company. The benefits that accrue to the company are among the most important factors ever identified as meeting commercial business objectives:

Two types of personalization technologies form the core of the personalization-centric effort: collaborative filtering engines and rules-based systems. Each addresses an important part of the business puzzle. There are many parts to running a business, but regarding customers, there are two: 1) knowing your customer and 2) conducting business based on customer knowledge. We will refer to "knowing your customer" as insight and "conducting business" as action.

Collaborative filtering is the dominant technology for developing insight and also provides the best starting point for action based on customer insight. Rules-based systems complement collaborative filtering by allowing action based on non-customer objectives, such as reducing overstocked items or geographic-specific marketing programs.

Start by Knowing Your Customer
Insight, the new term for individual customer knowledge, can be highly personalized with a collaborative filtering system. How does it work? The technology identifies electronic "neighborhoods" of individuals with similar preferences. The theory proven by collaborative filtering is: if Person One likes items A, B, C and D, then Person Two – who also likes items A, B, and C – is in the same neighborhood as Person One. Consequently, it is probable that Person Two will also like item D. Therefore, a collaborative filtering engine might reasonably recommend item D to Person Two. What makes collaborative filtering particularly powerful is that it generates these insights in realtime for each individual customer and learns over time to be even more accurate.

Expanded, the concept of neighborhoods based on preferences can encompass millions of individuals and millions of items. In fact, larger databases often deliver more precise insight. Also increasing the level of accuracy is the collaborative filtering feature of learning about individuals with each contact.

For example, CDNow uses collaborative filtering software to make recommendations of new music to its millions of customers from an inventory of hundreds of thousands of CDs. The system analyzes information about the visitor, such as prior purchases⁴, and draws from an electronic neighborhood a list of recommendations for that particular visitor. The collaborative filtering engine will update its understanding of the individual during each visit and instantly revise the neighborhood according to the new information. This system ensures that each new music recommendation uses all possible insight about the customer.

Consider for a second how revolutionary the development of collaborative filtering really is. In the past, greater customer intimacy was achieved best only in boutique companies such as the corner florist who knew you by name. Trying to achieve that level of intimacy at a large scale was nearly impossible. Now with collaborative filtering, high levels of customer intimacy can be achieved by companies that operate on a massive scale. In fact, the more customers, the richer the database of customer preferences, and the better the business can tailor its offerings to the unique needs of each customer individually.

Why Rules Alone Cannot Get You There
By contrast, rules-based systems need to make assumptions about their customers before they actually arrive at the site. Rules systems are based on marketing theories of segmentation and depend on carefully formulated categories of customers. Potential customers are segmented by a variety of factors, such as age, gender, location, education and income level. In addition to these demographic segmentations, there are often psychographic segmentations. More advanced systems segment customers by behavior such as most profitable customers, customers most likely to churn, bargain hunters, etc.

The rules need to be created carefully to ensure that all customers who appear on the site fit the market segments that were defined in advance. Those people who are mismatched to their market segments are less likely to receive accurate recommendations for products or services, and the business will miss an opportunity to personalize the interaction and delight those customers.

Consequently, a personalization-centric enterprise requires collaborative filtering to deliver the customer insight that fulfills its mission.

Meshing Personalization Systems
A successful personalization-centric enterprise wraps its business around the wants and needs of its customers. The greater the insight the business has about each customer, the more promptly and successfully it can fulfill those desires, thereby creating customer satisfaction with personalized experiences and nurturing customer loyalty with enterprise-wide knowledge and commitment to individuals. As we've seen above, collaborative filtering delivers the best customer insights.

A company's actions – the second step to delivering a personalized experience – must flow from these insights. In addition, each business will have secondary and tertiary goals that follow the number one goal of serving every customer individually. These other goals also require action and typically are well suited to a rules-based system. Rules systems are most useful when specific information is known in advance and consistent treatment is appropriate. Two examples of situations in which rules should be added to the customer interaction equation are operational constraints such as inventory and financial constraints such as profitability.

Consider the situation when a collaborative filtering system suggests a particular music CD that is on backorder. It makes sense to apply a business rule that recommends the second most valued CD for that customer and then flags the system to send a follow-up e-mail when the first choice arrives in stock. This combination of collaborative filtering and rules-based personalization serves the customer in an individual way by suggesting music he is likely to enjoy while simultaneously serving the business by activating a follow-up system to keep the enterprise from losing an additional probable sale. Of course, the business also benefits by preventing the customer's disappointment from finding the perfect CD and then being told it is temporarily out of stock.

To demonstrate a financial constraint, imagine a situation in which collaborative filtering recommends two music CDs that rate equally appealing to the customer. Perhaps the two artists are Ricky Martin and Christina Aguilera. A business rule applied to these recommendations might be to present the one with the higher gross margin first. Perhaps the Ricky Martin release has a promotion running for the month, so of the equally-ranked CDs, Ricky Martin would be recommended more prominently. An additional rule could be applied if the customer didn't buy both CD's to send an e-mail in a week or two promoting the second.

Attaching rules-based recommendations to collaborative filtering should be done with caution, however, in order not to defeat the personalization-centric focus. If rules are applied too often or inappropriately, customers may find that they are not receiving the level of service expected. If companies overly weight inventory and financial constraints, it is likely that customers will "smell a rat" and defect to competitors.

Optimized Personalization — A Summary
Collaborative filtering engines and rules-based systems are truly complementary technologies, as evidenced by the fact that Net Perceptions, award-winning leader in the personalization industry, offers rules-based capabilities to its collaborative filtering solutions. Businesses choosing to be personalization-centric will find collaborative filtering is the primary technology for developing individualized insight about their customers, and it provides the most comprehensive and accurate starting point for personalization-centric business activity.

Collaborative filtering is such a new technology that research is only now beginning to uncover the magnitude of the value it adds to a business's relationships with customers. Nielsen NetRatings, for example, recently found that sites using collaborative filtering dramatically outperform their non-collaborative filtering competitors. CF-integrated sites reported conversion of Web visitors into buyers at a rate two to three times that of their competition.⁵

Rules systems, however, have a role in the personalization-centric enterprise. Rules allow companies to take the insights of collaborative filtering and build their customer offerings in alignment with the constraints imposed by operational and financial considerations.

Grasping Change
Dealing with customers as individuals instead of groups is a new concept for marketers and may require a live demonstration. Personalization as a corporate ideology is also new and the only way to demonstrate it is to try it. As the ability to serve each customer uniquely becomes absorbed and applied throughout the enterprise, the instantaneous fingertip access to preferences and products will become second-nature. Manufacturing, order-processing, inventory levels, shipping, marketing, sales and customer service all flow smoothly in one direction due to the implementation of personalization with collaborative filtering.

Business as we have known it is changing. Personalization-centric businesses run more efficiently and effectively while simultaneously being more responsive to customers. This revolutionary change promises a one-to-one service economy that has never been experienced before and customer relationships that are elevated to extraordinary new heights.

Footnotes

1. (a) World Technology Award for Commerce, presented by the World Technology Network in conjunction with The Economist magazine. Founders Snyder, Miller and Riedl were judged among the individual leaders worldwide who most contributed to the advance of emerging technologies for the benefit of business and society.


(b) MIT/Sloan School of Management E-Commerce Technology Innovator Award, May 1999. The award was granted to Net Perceptions "for exhibiting the technological innovation with the greatest potential to further revolutionize Web-based commerce."


2. The Experience Economy: Work Is Theatre & Every Business a Stage, B. Joseph Pine II and James H. Gilmore, Harvard Business School Press, 1999. Appearing on Personalization.com.



3. The One-to-One Fieldbook: The Complete Toolkit for Implementing a One-to-One Marketing Program, Don Peppers and Martha Rogers, Ph.D., and Bob Dorf. Doubleday, 1999.



4. Customer information is gathered in a variety of ways, as defined by each site's business needs and Web site design. Information may be collected either implicitly (by observing customer activity) or explicitly (by questionnaires and surveys).



5. Nielsen NetRatings, November 1999.