Web behavior boosts customer intelligence

We present a new solution for personalizing web site interactions based on the fusion of customer and web behavioral data. This maximizes the value of company web sites and turns them into truly customer oriented channels.

“FinScore is a unique provider of insightful expertise in analyzing and predicting customer behavior and interests to optimize and personalize the user experience on our company portals”
-Erwin Roos, Head of Marketing & Sales Planning & Monitoring, Swisscom AG

Today, companies often keep CRM and web infrastructure/projects totally separated. CRM happens in the "big strategic IT corner" while WEB is rather placed in the "tactical techie corner". The respective departments do not talk to each other; they frequently do not even know each other. Since there is no "cross-talk" there are practically no synergy effects arising from a combination of the CRM and WEB worlds.

This isolation of both worlds is reflected by the fact that e-campaigns remain totally independent of ongoing CRM-driven campaigns. e-campaigns and their responses are rarely tracked in CRM-systems. This renders the measurement of end-to-end success of e-campaigns an impossible task. To make things worse: frequently a prospect may be contacted online, demonstrate interest in the offer and purchase over a different channel (i.e. in a shop and not online). There is no way to connect the event "purchase" with the fact that the lead had been generated via the web channel. This is bad news for those responsible to make their web channel business case as their knowledge is based on a narrow view of customer online interactions.

Customer data (CRM) and WEB data are not employed to enrich each other. This is far from optimal because, as we will explain later, huge leverage effects can be derived from each of these data domains. Currently, CRM or simply customer data which is available in most large companies is not used to enrich and enhance web user interactions and vice-versa web behavioral user data is not used to enrich customer profiles. As a result we end up with a limited customer profile, which enables only limited personalization of customer interactions.

Limitation of "CRM Only"

In the case where only CRM data are used to drive and personalize customer interactions there are severe limitations such as the absence of knowledge regarding customer level soft data (e.g. interests and affinities), which could be obtained from observing customer web site behavior. Either the CRM system does not contain this type of data or it is incomplete, of bad quality, purchased from address brokers, etc. Ignoring such customer soft data therefore results in poor enhancement of customer interactions and is very expensive. Also, in the case where the CRM system contains soft data it is not regularly updated and represents the customer at one point in time, i.e. at the creation of the corresponding customer record in the database.

As a consequence, we will end up with an incomplete customer profile since web behavior is totally ignored. Due to the static nature of soft data, even if customers change their behavior and interests over time, there will be no reflection through an adequate time-varying personalization of the interaction. In summary, working only with CRM data for the web channel is making very inefficient use of it as a channel to reinforce customer relationship, acquire new customers and generate new leads.

Limitation of "WEB only"

Only using web data is often called "Behavioral Targeting" (BT). The main challenge for this approach is that it bases its decisions on anonymous visitors and has a very session oriented view. As a result, only short term web profiles can be built, leading to a very volatile impression of the visitor's web profile. In reality, people's needs are rather stable over time, apart from some inherently short term interests (e.g. when browsing the web for your next vacations). Recommendations for personalizing the web visitor interaction are therefore not addressing the "real" individual visitor's needs, but are blinded by short term effects. This restricts BT to applications where offered products, services and content address typical short-term needs.

A new solution: fusion of CRM and WEB data

Objectives of this new approach are twofold. On one hand, to personalize web interactions by delivering the right messages, products, banners, etc. to the right web site visitors. On the other hand, to generate new customer intelligence based on visitors' web behavior.

By adopting this approach, the impact of the company web-channel will increase. A lot more will be learned from customers by taking into account their web behaviors. It turns the web into a truly personalized channel to generate leads, acquire new customers, cross-sell to existing customers and retain them.

Recommendations for making the visitor experience unique are generated for three types of web site elements. Redactional content of a web site can be taken as an element to target information presented to a specific visitor. This includes categories such as news, product information and updates, and help topics.

Another element for personalization is, especially in an e-commerce context, targeted placement of products or additional services: present those products most likely to interest the visitor in order to increase customer revenue or special additional services such as online billing in an effort to reduce customer related costs.

Finally, personalization is also driven by placing the right advertisements (banners) for the right visitor to increase conversion rates, maximize click-through rates, and optimize the effect of branding campaigns.

Who is it for?

This solution has been especially designed for companies already disposing of a reasonable amount of CRM data, owning one or more web portals not only for promoting their products and services, but also for offering web based self services such as online shops or online banking. Companies likely to profit from the presented solution are banks, telecommunication providers, insurances, retail shops with a web based sales channel presence, and pure e-commerce providers.

What does it do?

The solution addresses four scenarios for web personalization.

CRM & WEB-light – Based on identified visitors where CRM data and very limited web data such as banner clicks are available, a series of recommendations can be generated for future web site visits of customers. This drives the web site personalization in terms of targeted content, product offers and ads.

WEB profile completion – Based on identified customers where CRM and web data is available, model based predictions of web interest affinities can be generated for customers who have not yet visited the web site. This enables personalizing the web interaction from the very first web site visit.

Customer (CRM) profile completion – Based on identified customers where CRM and web data are available, additional CRM attributes such as gender and age can be predicted for web site visitors who are not customers. This helps understanding socio-demographic properties of the web site audience in order to provide adequate communication, ads, and content.

CRM & WEB Fusion – Based on identified visitors with CRM and web data, a series of recommendations as well as web interest affinities (e.g. sports, technology, product type A, product type B) can be generated for customers with repeated web site visits. This drives the web site personalization in terms of targeted delivery of content, product offers and ads.

How does it work?

We distinguish two types of web site visitors: identified and anonymous. For identified visitors, both CRM and WEB data exist. For anonymous visitors, only WEB data are available. Usually, the majority of website visitors are anonymous.

The principle behind the presented solution is to learn from identified visitors to infer powerful recommendations for all web site visitors. This increases customer knowledge, targeting reach, and quality of website personalization.

We collect web behavior data for identified web site visitors and combine them with available customer level data (CRM) to infer unknown CRM attributes, WEB behavior, and recommendations (content, products, ads) for personalizing the web experience. The solution thus augments customer profiles, generates customer intelligence,

Given that web data are huge, it is a demanding task to process and aggregate to a level where data volume is minimized while maximizing the web behavior information content. The solution's web data aggregation process takes care of this and summarizes data, based on page views, into customer level data. It generates a rich web profile for each visitor with hundreds of web data based attributes such as the amount of pages consumed by content type during the week, weekend, in the morning, at night, etc.

Once the web profile has been generated for each web site visitor, it is merged with available customer data (CRM) for all visitors also being customers. This leads to an extended customer profile consisting of socio-demographic attributes, information about the customer product portfolio (which products does she own?), product usage (how are the different products used?), and web behavior profile.

The next step is the automated generation (usually in offline mode over night) of a predictive recommendation model for each item being personalized for the web site interaction (e.g. redactional content, product offers, ads).

Each recommendation model is then used to generate personalization scores for each visitor depending on the chosen scenario (as described above).

The recommendations are made available to the web site and content management or ad server systems to finally drive personalization. As a further feature, the proposed solution allows to adaptively learn based on daily web data. This means that for each new item to be included in the personalization, the performance increases daily.

How is it implemented?

To implement the proposed solution, one needs to integrate it with the targeted web site and customer data repositories (typically a CRM system or a data warehouse).

For the web site integration, a few requirements exist: access to daily web log files must be guaranteed. They are generated by various servers attached to the website: web server log file tracks page impressions with information such as the URL (page) visited, timestamp, browser and operating systems. The ad server (e.g. DoubleClick) log files track which ads have been shown on a specific page and which ads have been clicked by the visitor. e-commerce server log files track which products have been presented and purchased by the visitor/customer.

In order to build up web behavior profiles it is required to know for each URL which content category is associated with it. The content of the targeted web site needs to be categorized either by providing a mapping of the site's URLs to content categories such as sport, finance, news, games or by tagging the pages via a content management system. Also, a list and categorization of items to be used for personalizing the interaction is needed. This list contains items depending on the specific context of the implementation such redactional content, products or ads along with their respective categorization. This allows linking a specific score with a recommended item (e.g. show products of type X or a banner of type Y to this visitor).

Regarding the customer data repository integration a pre-defined set of customer attributes needs to be made accessible to the system. Attributes are divided into three categories: basic customer data, product ownership, and product usage. Basic customer data includes typical attributes such as age, gender, city, zip code, tenure, household size, marital status, etc. Product ownership includes attributes such as amount of products purchased per product or product type, date of purchase, price per purchase. Product usage includes attributes such as frequency of use per product or product group, volume per use (e.g. in telecommunication the average minutes per call, most frequently called countries, monthly billing).

What are the benefits?

A multitude of benefits arise from applying the presented solution to personalize visitor interaction with company web sites. This includes increasing banner click rates (recent online advertising campaigns increased click rates by more than 400% when optimized with our solution), more product sales through targeted cross- and up-selling, increasing brand value due to targeted delivery of promotional messages to the right audience, pushing lead generation up through increased web site stickiness, and adding important value to increase customer intelligence through fusion of customer level web and customer intelligence.