Technology-assisted, hands-on analysis of business data goes back at least 30 years. The terms describing this activity have changed as business needs evolved and as technologies advanced, but the core essence of the practice has remained the same.

Analytic services gave way to end-user computing as fourth generation languages (4GLs), and the first wave of dimensional tools replaced hard-core statistical packages on the mainframe. End- user computing morphed into decision support while client-based graphical tools took hold and data stores proliferated. Decision support was subsumed under the more ambitious label of business intelligence as the second wave of dimensional tools (now called OLAP) joined the growing plethora of reporting and analytic options.

Then the World Wide Web changed everything.

Technology reversed to a network- and server-centric model. Every vendor fought to catch up and one-up the competition by defining a new paradigm. All tools had to become Web-enabled until that was not enough, and everything had to go native.

But, the real change was in the nature of business itself. E-business is the more mature outgrowth of the dot-com frenzy, and it’s here to stay. It is fundamentally different in several ways,with many more differences yet to be discovered.

  • How will the practice of business intelligence change to support this novel, volatile, evolving and unpredictable business model?
  • Will the seemingly seat-of-the-pants, cowboy mentality of e-business change with the incorporation of this more left- brained activity? Is this good or bad?
  • Is it even fair to cast this new form of analytic activity in such a geeky fashion when more intuitive and creative elements are incorporated?

I can only scratch the surface of these questions. Let’s start by defining the unique aspects of e- BI.
Business intelligence for e-business has several focus areas: economics, exploration and enablement. Economics involves implementing the (often missing) infrastructure to capture and track performance data. Exploration entails a discovery process that interrogates all data captured via interaction with the site, participant profiles, partner data, market information and other external factors. Enablement involves deploying dynamic processes to create new business opportunities.

Economics

The dot-coms that survive will be those that build a viable economic model and manage their businesses accordingly. Entrepreneurial zeal gets the companies started. Operational effectiveness will ensure their long-term success. The practice of e-BI could start with the basics of performance tracking.

The end goal is to initiate momentum toward profitability. This is what the market is demanding now. Building a market and locking in customers is not enough unless costs can be managed and revenue can aggressively grow.

The dot-coms have the advantage of minimal infrastructure (no out-of-sync legacy applications) and relatively simple monetary flows. This makes it much easier to implement tracking mechanisms once profitability targets are set. Well-established techniques can be applied that leverage traditional BI skills. We have to build from scratch, but in very familiar territory.

Introducing economic accountability into the Internet ventures of a brick-and-mortar company is far more difficult. It is difficult to segregate the revenue streams and expense accounting of the e- business from the existing business. Often, shared costs are assumed by the traditional business, and an overly aggressive allocation of revenue is made to enhance the apparent success of the e-business wing. Spinning out the e- business division is the only sure way of creating divisibility of economic results.

As long as the e-business is still within the existing corporate structure, e-BI is more difficult – but it is also more essential. A detailed operating model must be constructed that goes beyond simple allocation of costs and revenues. Transaction by transaction, costs must be attributed to the old or the new business. Where the costs are truly shared, they must be tracked separately to see how material they are to the profitability of the e- business based on how much of the cost is allocated.

At first, it seems that revenue should be easy to apportion. Revenues from Web site advertising and transactions go to the e-business, while retail store revenues go to the existing business. But, it gets murkier as you look deeper. How do you handle cross-promotions where the Web site encourages increased retail traffic such as Web-delivered retail coupons? How about when retail outlets sell low-margin equipment while referring customers to the Web site for high-margin (but high- inventory cost) supplies? The role of the e-BI process is to detect, monitor and analyze these differential impacts on profitability.

Exploration

An e-business throws off a lot of data – clickstreams, inquiries, orders, support requests and interactive contacts, just to name a few. These can be very rich data streams because we get the who, what, when and where of each event. We can also get how long, how much and what for information in some cases. Even more importantly, we can get the entire life cycle of interaction with a given customer from a single source. This has been virtually impossible with traditional businesses that touch a customer in many different ways from multiple departments on diverse and nonintegrated systems.

The e-BI challenge is to make sense of this rich, but raw, base of data. The first essential task is to clearly articulate all your touchpoints with the customer. You build a lifecycle model that accounts for all the stages of interaction from first sign up to account termination. Within the overall customer life cycle are buy/interact cycles. These repetitive cycles have stages that apply to each product from inquiry to order to delivery to support requests to returns. Other cycles apply to other forms of interaction.

The purpose of life cycle modeling is to ensure we can chain together all the data for a customer and each sequence of interaction. It would be a shame to finally have an environment that promises holistic and integrated information but fails to deliver because of business process or data-design oversights.

The second step is to build a dimensional model using published and proven methods. The dimensional model helps monitor progress toward defined targets within a formal analytic structure. It can also be used to identify and set new targets.

One essential variation is to build from the detail up without the disconnection that results from intermediate aggregation. Our first implementation is always a detailed star schema to preserve maximum flexibility during initial discovery and dimensional refinement.

When you have identified repetitive business processes, you should invest in an OLAP tool of your choice. The introduction of this technology is primarily for performance management and to provide enhanced, out- of-the-box functionality for analysis. It is important that you not move to implement OLAP tools until you are ready to lock and load. You must first achieve a degree of stability in your dimensional design, and you must understand the business processes you intend to support. The introduction of the tool provides many advantages, but it will definitely restrict your flexibility once implemented.1

The greatest potential can be realized using data mining techniques. One of the promises of e-business is to analyze the patterns of customer interaction to identify marketing opportunities. Data mining tools are just now coming into the mainstream. The limiting factor will not be the technology; it will be your imagination.

The simplistic view of data mining is that you pour undifferentiated data into a cauldron, stir as it boils and bubbles, wave your hands over the brew and, voila, a conjecture materializes. Unfortunately, data mining is hard work despite how easy the tools may be. You must define an analytic framework and select the decision variables. This is where imagination comes into play. If you do not collect and include the right data, you may get inconclusive or, worse, incorrect results.

For example, consumer commerce sites would like to encourage impulse buying. They do this by displaying targeted offerings. Without data mining, a basic technique is to offer you items similar to what you have already bought. If you have ever bought Pokemon cards, the site may offer more Pokemon material or other children’s toys. This can be very annoying to someone who bought these cards once for a nephew and does not regularly buy for children.

Now, say you use a data mining tool and include "number of children" and "number of purchases per category" as variables. The tool is likely to predict that people with no children will generally buy fewer toys. Duh! How do you identify people who are more likely to buy toys online and respond to an impromptu solicitation?

What if you add marital status to the mix? Okay, in advance you expect that married people with children obviously buy more toys. When the results are in, you might find that divorced parents buy more online than married parents. This is a more meaningful result. Including sex might reveal, as one company found, that divorced fathers are far more likely to purchase toys online than any other segment by a large margin. They are better candidates for online promotions.

When the results are in, many conclusions from data mining seem self-evident. But you can’t prove them apriori. You will also never find these relationships unless you have the foresight to feed the data mining engine the right factors in the first place.

Enablement

The final category, enablement, refers to exploiting business intelligence methods dynamically as a part of doing business. BI is no longer just an optional, adjunct analytic process. It becomes a method of defining, tracking and revising new programs and new business opportunities.

In my May 2000 column, the topic of closed-loop analysis was introduced. This is an aspect of e-BI that supports just-in-time decisions and continuous revisions. This is a tightly coupled process where analysis spawns execution, and the results of executing a tactic become fodder for follow-up analysis.

A meaningful example is deploying an active customer retention program. With many traditional businesses, growing customer share is far more valuable than gaining market share. It is more cost-effective to sell more to existing customers than recruit new ones. This is a hard-won lesson that is only now coming to the attention of Internet enterprises.

E-business has followed the example of telcos and ISPs by aggressively targeting new customers. Dominating a market niche has been the mantra of e-business. This tactic has worked for a few lucky ones, but at extremely high cost. Reducing churn – the measure of customers coming and going – is not just a business goal, it is a survival tactic.

What will keep customers coming back to your site? You have two options, expand or enrich, and both are enhanced by e-BI. E-BI can be exploited to reduce the cost of expansion into new markets, but this is more of an intuitive than data-driven process. It is far more effective at tracking progress to confirm a success or to cut your losses early.

Enriching the experience of existing customers is far more amenable to e-BI integration. Enrichment involves adding new features and services that set your site apart from all others. This is an incremental process of continual refinement. Targeting, testing and tuning the new offerings allows you to both generalize and customize. You generalize by identifying those features that appeal broadly to existing customers and help attract new ones. Customization entails identifying narrow segments of your customer base that respond more favorably to a particular new facet.

E-business is the first business sector to live and die by information presentation. Once the infatuation phase is over and customers become more discerning, business intelligence will become the dominant profession of the Internet information age.

We will explore e- BI in more detail in future columns.

Return to Reason

By Michael Haisten

Business intelligence for e-business will grow substantially despite the stock market recoil from dot-com companies. In fact, the return to rationality will ensure the viability of this market for the foreseeable future. E-BI serves the broader e-business sector, not just the dot-com start- ups.

In a nutshell, the future finally looks bright for this sector. Things looked really bleak during the last few years as hype overtook reality. The market was intoxicated with any company with an "e" or a "dot" in its name, proving once again that the herd instinct can overrule the invisible hand of rational economics in a free market. I was reminded of the stark commercial from Apple Computer in 1985 showing legions of people following one another over the brink of a sheer precipice. Just change the implied Pied Piper from IBM (the Microsoft of 1985) to a dot-com in order to convey the new lemming syndrome of the late 1990s.

When anyone with a half-baked idea (touted as a vision) and a cheesy piece of software is showered with tens of millions of dollars; all reason has been abandoned. Many of the business plans for these start-ups were devoid of any practical business goals. Both the founders and their financial partners focused on an exit strategy to make them rich rather than a sustaining business strategy to help them live long and prosper.

The collapse of the dot-com frenzy heralds a return to a focus on fundamentals. The practice of e-BI is poised to help those in e-business with a real plan of action.

  1. Regardless of vendor rhetoric, you cannon have high flexibility, high functionality and high performance in a single implementation. This is not a product flaw; just a fact of life. These factors are all relative and form a classical trade-off. The role of OLAP tools is to optimize performance with high functionality. An essential aspect of any dimensional model is that it represents a specific interpretation of the information context. It is, by definition, inflexible.

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