The explosive use of e-mail is offering brave new opportunities for eliciting unique customer insight from the intelligent analysis of e-mail. The massive growth of e-mail is a result of its unique ability to bridge time and space in conversational communication for pennies. Phone or print mail cannot match e-mail's accessibility and ubiquity. To effectively leverage business intelligence (BI) for e-mail, it is important to gain a clear definition of e- mail and the unique role it plays as a medium of communication.

Pattern analysis is a foundation of BI and has significant applications to e-mail; however, application of pattern analysis requires a fundamental understanding of the properties of e-mail. In this article, we explore the unique properties of e-mail that are central to any BI applications for e-mail.

Many organizations are already using BI to extract insight from e-mail; while most such applications are rudimentary, organizations are just beginning to realize its importance, particularly when they discern the vast amounts of tacit knowledge resident in e-mail dialog.

What is E-Mail?

E-mail is clearly the fastest growing communications medium in the world today. More than 5 billion e-mail messages are transacted each day, and human beings are engaged in the activity of writing as never before. Centuries ago, only the elite or scribes were allowed to write and disseminate their thoughts and ideas. E-mail has fundamentally removed such age-old barriers to writing, enabling millions to dialog and discourse in a one-to-one, high-velocity manner, unparalleled in the history of communications.

Friends who found it hard to stay in touch can do so on a regular basis with relative ease. The medium offers an individual time to formulate thoughts and feelings in a personal environment. Customers who previously had to wait on the phone all day for a customer service representative can find the e-mail address of the CEO and e-mail him or her directly.

E-mail has become a part of everyday life. It is here to stay.

Today, we all use e-mail; some of us spend our entire day corresponding via e-mail. U.S. corporations now receive nearly 50 million inbound customer care e-mails per day. Fortune 1000 companies are budgeting two to five percent of their traditional advertising budgets for e-mail marketing.

In 1977, my first definition of e-mail was: electrified, transportable paper. My more mature definition of e-mail in 2002 is: an electronic medium for asynchronous conversation.

The three key elements to this definition are: medium, asynchronous and conversation.

A medium is said to be a substance through which something else is transmitted or carried. Asynchronous means that an event is not synchronous or not occurring or existing at the same time or having the same period or phase. Conversation in this definition means the exchange of thoughts, opinions and feelings. Thus, e-mail is a medium through which thoughts, opinions and feelings can be exchanged with another, without the need for either party to be present at the time of the exchange.

How is this different from paper mail? The substance of the medium is electronic versus paper, thereby offering infinitely far more flexibility than paper. This flexibility allows us to forward, broadcast, carbon copy, confirm receipt, etc. The ease and low cost of e-mail are not features of paper mail.

E-mail also has some very unique properties that can enable sophisticated BI.

Properties of E-Mail

In physics, matter has properties such as density, weight, Young's modulus and Poisson's ratio. By understanding these properties, modern engineering is possible, enabling us to manipulate matter in ways unimagined. Analogously, foundational understanding of a medium is a prerequisite for engineering that medium effectively. The understanding of the properties of e-mail is central to effectively creating unique BI applications to leverage the medium to gain analytical insight.

What are the properties of e-mail? A story will serve to illustrate the properties of e-mail that I discovered in 1994.

In 1993, the United States White House of President Bill Clinton was receiving approximately 5,000 e-mails per day, pre-World Wide Web (WWW). Approximately twenty student interns would read these e-mails and manually categorize each e- mail into one of 147 different issues, established by the Executive Office of the White House. The issues included categories such as: education, health, drugs, threats, Hillary, etc.

Each week, the Executive Office would receive tallies and analysis of volumes of e-mail by the different issues. These reports provided a pulse of the Clinton constituency from the BI gleaned by manual sorting of e-mail. Clinton was ahead of his time. He was leveraging e-mail to listen to the thoughts and opinions of his constituency.

In 1994, the White House decided to find ways to automate this manual process of e-mail intelligence gathering by sponsoring an industry- wide competition to find technologies that could automatically analyze e-mail by issues. At that time, I was a graduate student at MIT focused in the field of pattern recognition, developing pattern analysis techniques for recognition of complex ultrasonic signals. Fortuitously, I became involved in this competition because of the advice of a professor.

At that time, many search engine companies and text-mining companies were leaders in this field; however, most of them took a very natural language processing (NLP) or text- based approach. My approach was a pattern recognition signals-based approach using a variety of hybrid methods in feature extraction, clustering and learning. More importantly, pattern recognition is a discipline that has evolved over nearly 100 years. In that discipline, key to success is identifying properties or features.

After many sleepless nights of reviewing hundreds of thousands of e-mails, it was clear that e-mail too had some unique features and properties. From my empirical research, e-mail has the following properties: attitude, issues, requests, objects and sender type.

By using these properties, I layered on signals-based techniques in feature extraction, clustering and learning. The result was 78.2 percent accuracy of automatic categorization. This resulted in the winning approach to the White House competition.

Each e-mail may have one or all five of these properties. These properties are important because they give a rich insight to the individual behind the e-mail. When you or I write an e-mail to a company, we express an attitude; remark on issues of concern; potentially ask for resolution through requests; and invariably, in that same e-mail, give an indication of what object or product is the source of concern. Finally, in most e-mail, individuals also share intimate details about themselves.

For example, a typical consumer e-mail may read as follows: "Dear Company X, I am very upset at the fact that the shoes I bought squeak and would like a refund or replacement. In addition, please send me the annual report. Also, my son is graduating from college and would like to work in your marketing department. I love running and football and have been a great fan of your company; however, the recent problem has me quite disappointed. Help!"

In this example, the individual reveals his thoughts, opinions and feelings. The writer is sharing his attitude (negative), his issue (squeaky shoes), his requests (job application, annual report, refund), his object (shoes) of concern, and finally he shares his personal interests or sender type (running and football fan).

Recognition of the properties of e-mail serves as a foundation to extract customer insight from e-mail analysis.

Business Intelligence and E-Mail

Intelligence agencies of the government have been applying BI to analyzing phone calls, trash, e-mails, etc. to avert calamities and gain insight into future behavior patterns for many years. These organizations clearly understand the power of using BI to leverage such intelligence to save time, money and lives.

What if Firestone had understood the power of BI for e-mail? There must have been at least several customers complaining about tire problems in their e-mails long before the major crisis hit Firestone. No doubt, they expressed certain attitudes and issues relating to tire failure and, more than likely, also had requests relating to alleviating their concerns. One can argue that BI effectively applied to their customer e-mail, at minimum, would have given some employee reason to voice concerns internally or raise a red flag – potentially avoiding the millions of dollars in loss from the legal claims.

The district attorney of New York, however, did leverage e-mail well. Through the analysis of Blodgett's e-mail correspondence and dialog with other employees and bankers, he effectively gathered intelligence and evidence on how Henry Blodgett, lead analyst at Merrill Lynch in the dot-com era, paid lip service to his own public buy recommendations and defrauded investors. Such analysis of e-mail led to a full exposure of the clouded lines between research analysts and bankers in the high-stakes game of investment banking.

These examples illustrate the power that e-mail analysis can afford organizations. Given the explosive growth of e-mail, organizations have much to gain by applying BI to their own massive volume of e-mail to extract knowledge and wisdom of customer behavior. If one believes in the age-old adage, "The customer knows best," then it behooves organizations to apply BI to e-mail. Such BI can help to gather additional tacit information on customer needs which can be combined with traditionally structured demographics and transactional data to build a holistic view of the customers' needs in real time.

The Challenge

One of the main challenges to fully deploying e-mail BI applications, however, is the confusion created over the past three years from "be-all" and "end-to-end" customer relationship management (CRM) solutions. Organizations, through their own bitter experience, particularly since the dot-com fallout, are now awakening to the reality that specialization of applications for managing a particular channel is the right model versus attempting to unify channels in "blended" model. It is becoming more clear that when one tries to make a CRM system do everything (ERP, data warehousing, e-mail management, billing, telephony, chat, etc.), one gets something that does nothing well. This realization is leading to unification of data, but specialization of channel.

This is good news! This trend is allowing us the freedom to honor e-mail for the medium it truly is and is helping us to recognize that e-mail as a specialty requires unique processes, strategies, workflow, technologies and BI applications specifically created for e-mail management, unique from other media.

Register or login for access to this item and much more

All Information Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access