Intelligence technologies that e-business demands

In today`s dynamic e-business environment, superior knowledge is often what distinguishes a company from its competitors. Willie Bezuidenhout, business technologist (Information Management) at Computer Associates (Africa), looks at two key intelligence technologies - Automated Rules and Adaptive Pattern Recognition - that enable e-businesses to leverage superior knowledge for competitive gains.

In the age of e-business, superior knowledge is often the only differentiator in the highly competitive marketplace. To be successful today, electronic organisations need to acquire, process, retain and distribute knowledge throughout the enterprise, making it immediately available to decision-makers at the operational level.

With a diverse set of challenges surrounding e-business enterprises, the appropriate technology must be applied in order to achieve universal goals - to grow the business, gain competitive advantage, and increase customer loyalty.

Many decision-makers will testify that without the use of intelligence technologies, e-businesses have little hope of an accurate understanding of their true marketplace positions and what strategies and techniques should be adopted to beat the competition.

To realise even a fraction of their potential in a largely conventional marketplace - never mind the complexities of competing with similar electronically-oriented organisations - e-businesses need a broader and sharper focus, a better ability to "see complexly".

Two key intelligence technologies have emerged for the e-business. They are Automated Rules and Adaptive Pattern Recognition. Ideally they need to work in tandem at the operational level for the best results.

These new generation intelligent e-business solutions not only analyse an organisation`s history and its current position but, more importantly, are able to formulate a strategy for future success.

A typical automated rules-based system stores, manages and executes several hundreds or even thousands of rules that are inter-referenced and linear. In its simplest form, a rule is a conditional statement with a well-defined outcome.

For example, if Tom Smit is interested in computers, then send him subscription offers from IT magazines. If Joe Brown is interested in cars, then motoring media subscriptions and offers should be targeted at him.

Business logic for an e-business, however, is more complex.

In an e-business, information is stored as a collection of inter-related declarative statements, which are separate from the business process code.

These rules are not stored as nested "if - then - else" statements, but as freestanding conditional statements. This allows for easy propagation of changes made in one rule to the other related rules without having to recode the entire rule base.

An "inference engine" takes in initial system conditions and accesses the rules base to derive a logical outcome in accordance with the specified rules.

For example, in a fraud detection scenario, the decision as to whether a particular transaction is fraudulent depends on the results of several freestanding rules that need to be linked through the inference engine.

These include the value of the transaction, the experience levels of the people involved, the stature of the players in the business community and the risk factor based on the exposure to various influences and media - such as the Internet.

Decision tables are built to map each input to its possible outcome(s).

Obviously, the path taken by the inference engine from the input to the output depends on the result of the rule at each step.

Depending on the application, the output of an automated rules-based system could be a bill of materials, an insurance detail sheet, or any other structured document.

Remember the days when you could walk into your local butchery and the shopkeeper would know exactly what you wanted? The kind of meat you preferred, the cuts and amounts to suit your budget, and most of your other preferences, seemed like common knowledge.

No doubt the shopkeeper also knew the store`s inventory by heart, thus knowing which cuts to cross-sell. Of course, knowing exactly what you already had and what you lacked provided a premise for up-selling. It was good business indeed.

How can you provide a similar customer experience in the online world?

Many businesses are looking into various techniques to personalise their e-business. One of the best ways to achieve this is by using adaptive pattern recognition technology to emulate human intelligence.

By "learning" relationships in data, adaptive pattern recognition technologies can predict future behaviour, make suggestions, cluster the appropriate entities according to degree of similarity, or optimise business conditions.

Adaptive pattern recognition technologies can provide an e-business vendor with a broad view of customers, offering insight into who they are, what they have purchased and what they want at a particular moment in time.

For instance, an online e-business vendor can profile the current customer based on income bracket, family size, profession and geographic location and suggest additional products for sale.

Along with profiling, adaptive pattern recognition technologies also enable site personalisation. Personalisation allows your customers to find what they are looking for quickly, without having to wade through a large amount of irrelevant content.

In today`s competitive markets, dynamic personalisation is not merely a convenience; e-business customers demand it.

The amount of data companies are collecting is increasing at a phenomenal rate. How can we effectively leverage this vast amount of information to create business intelligence and thus competitive advantage?

By analysing the large data repositories that corporations have collected, adaptive pattern recognition technologies can learn inherent relationships and create reusable intelligent models.

An example of adaptive pattern recognition technology is the neural network.

The term "neural network" is used to describe a body of algorithms that simulate learning and storing experience from data. A neural network is a mathematical function that computes the value of the output of some process based on the values of inputs to that process.

In order to effectively implement intelligence at the operational level, it is necessary to address both the logical and intuitive domains of business. Since automated rules are used to capture known business logic and expertise, and adaptive pattern recognition technologies are used to provide predictive capabilities, combining these technologies will deliver advanced, comprehensive e-business intelligence.