In the second part, I'll cover the remaining two major technologies for customer relationship management (CRM): rules-based personalization and advanced analytic and data mining solutions. I'll also touch upon how businesses are starting to realize (once again) that personalization isn't just a technology, but a full-blown set of business and technical strategies to recognize, understand, and fulfill individual customer desires.
Its most effective use, however, is with complex transactions, such as loan origination. In these transactions the seller has clear, well defined practices in identifying potential customers and how it needs to approach them. Rules can be used to streamline and automate otherwise lengthy and tedious underwriting processes.
A big advantage of this technology is that, unlike click-stream technologies such as cookies and collaborative filtering engines, business rules can be adopted for use in any platform, and thus are not limited to the Web.
The disadvantage is that rules-based personalization assumes that you understand what the customer wants and how she wants it. The solutions are also not very scalable, since the rules need to be constantly maintained. Having a well-defined process for re-setting the rules helps, and users can modify rules by providing explicit preferences or commands.
In this last category, I've lumped together all the predictive modeling and data mining software and systems that businesses can use to develop insights and uncover hidden relationships in customer shopping trends not readily discernible by traditional metrics (such as repeat purchases).
Before we continue, let's start with an acknowledgment: behavior analysis -- using algorithms, trending, decision trees, samplings, parallelisms, and other techniques to uncover patterns and predict likely behavior -- is still as much art as science. Consumers are complex individuals who don't readily fall neatly into the categories statisticians create for them. Nor can their behavior be readily predicted based on the past.
As you may suspect, data mining strategies fall in a wide range, including using simple database commands, selecting a subset of the data to analyze, and drilling into the data and identifying results with sophisticated analytical techniques.
These state-of-the-art analytics include statistical tools such as CHAID (Chi-Squared Automated Interaction Detector -- a market segmentation technique) and CART (classification and regression trees); smart data mining applications that use algorithms such as Bayesian networks to analyze and predict demand; and online analytical processing engines to crunch the numbers.
To the extent businesses can get the right data (by asking the right questions) and set up their data warehouse accordingly, this will dramatically reduce the investment required to produce meaningful insights and actionable information.
If there's a common lesson to be learned from the three major CRM technologies -- Web analytics, rules-based personalization and advanced data mining -- it is this: traffic-pattern statistics and pattern-recognition algorithms only get you so far. That's because the real trick is to discover what customers want now, which often can't readily be determined by analyzing the past.
While some of these technologies have found a useful role, more than a few have not. Many businesses have discovered that context -- for example, the need to discriminate between purchases for oneself and those meant as gifts -- is key. An officiated example is the strange recommendations a collaborative filtering engine generates to a classical music buff after he buys an Eminem CD for his nephew's birthday.
Rules-based personalization has its place. But current applications tend to provide simple, cosmetic personalization, on an anonymous basis. Many of the so-called personalization (both rules-based and collaborative filtering) products such as Art Technology's ATG Dynamo, Blaze Software, BroadVision, and Manna are used to serve targeted content, which is not exactly a high-impact sales proposition. True personalization is dynamic, proactive, knowledgeable, responsive, and personable.
And while data analytics can be impressive, they have limited value without a clear data management strategy -- i.e., a way to determine which information is useful or predictive, and a way to capture and manage the data to get an integrated view of the customer. In the absence of a data management strategy, most businesses simply decide to store all information, which, due to declining storage memory costs, seems a safe way to go.
The Importance of Relevance
Relevance is the single most important aspect of personalization, and it doesn't come cheap or easy -- which is why the focus of this investment should be made on your biggest and best customers.
The ability to provide a relevant offering to an individual customer calls for a mix of technologies (including the good old-fashioned technique of simply asking customers what they want instead of trying to divine the information using neutral networks). It also requires a supporting infrastructure of people, processes, and tools. After all, there's no economic benefit to discerning a customer's interests if you can't fulfill them.
This means not only building the back-end fulfillment capabilities to meet demand, but also providing customers incentives to stay loyal. For example: "You recently bought something from us; we'll give you a discount on all purchases for the next few weeks." Or, "You haven't been here in a long while. Welcome back and please accept this gift with your next purchase!"
Discovering who your customers are and what makes them tick is a major undertaking. So you need to ensure that potential payback is worth the investment.
This should begin with an understanding of customer lifecycle value, which, by the way, can be dramatically impacted by loyalty strategies. Studies have found loyalty programs can not only dramatically increase (sometimes even double) retention rates, but also increase the amount a consumer spends.
On the technology front, CRM companies are developing increasingly integrated approaches. Sellers of analytic applications are partnering with campaign management and e-mail marketing vendors to provide more targeted and complex technologies, dynamically generated and tailored to specific users in real time.
But we must remember that personalization is not a technology. It is not about marketing. It is not even about selling (or cross-selling or up-selling). It's about building and maintaining a relationship. It's developing the ability to know what any particular customer wants, and when, how and why she wants it -- which in turn is all about strategy, tactics, business processes, and understanding your customers.
It's amazing how many businesses whose competitive strategy is to differentiate themselves by getting closer to customers still treat all customer touch points as opportunities to sell them more of something. These organizations need to realize that repeat purchases and cross-selling are a result -- not a cause -- of customer loyalty and satisfaction, which is driven by the quality of the relationship between business and customer.
Organizationally, most businesses are still structured to treat customers as transactions, not as individuals. Sales forces sell; call centers handle complaints and requests.
Businesses are learning (and re-learning) that relationships, like communications, are two-way. Companies who seek to build and sustain committed relationships need to view and treat customer interactions as critical opportunities to listen, learn, and add value.
Arthur O'Connor is a senior manager in the financial services practice of KPMG Consulting, specializing in customer strategy as well as related architectural and organizational issues. An author, speaker, and consultant, he focuses on customer relationship management and eCRM.