DATABASE STRATEGIES: Integrating Marketing Databases With the Web

We’ve been approached lately by many clients who’d like to know whether their marketing databases can be integrated with the Web.

Integration, however, means different things to different people. You might have a direct response agency working on your marketing communications strategy through the integration of customer data in a combined online analytical processing and data-mining technology environment. They obviously are not all doing the same thing.

When we talk about linking marketing databases or customer relationship systems with the Web, it’s first helpful to understand the concept behind it. Integration of traditional marketing systems with the Web could mean:

– Source-to-base data transport

Data collected on the Web site or via another e-commerce application can be readily ported into a customer-centric database or central data store that includes all the data collected and generated from more traditional media.

– Data analysis

Information derived from data analysis can be used in future channel and marketing strategies, particularly Web strategies. You might find that visitors to your Web site aren’t purchasing anything that doesn’t have a direct link to the home page. You might also discover demographics of your online customer that would drive your Web merchandising and advertising plans. Some companies are now building profiles and contrasting between their e-customers and traditional direct mail or retail customers. These firms are able to draw comparisons between the activities, current value and related revenue potential of those who shop using a paper catalog, at the store, online or using some combination of these channels. Such data integration can be relatively easy to accomplish, and obviously can be very valuable.

– Application sharing

The analytical tools developed and used in a traditional marketing systems environment – such as lifetime value projections and response models – can also be applied to the electronic base of customers. Traditional selection techniques and testing approaches can be applied to the new channel as well.

– Base-to-source data transport

Data maintained in the marketing system can be used to populate a separate Web-based database. The Web server may be able to maintain limited customer data that can drive customized or rule-based content. For example, IDs and passwords are used on many sites to enable a simple approve/reject decision process. IDs may also determine whether you’ll link to all areas of the site or a limited subset of content. Alternatively, some high-tech companies will use this information to create semi-customized site content. In all cases, some small set of data is maintained on the Web server, apart from any broad-based collection of historic customer information.

– Data reaction

Many interactive Web sites are driven by rules-based decision processes.

The approach has been to establish large “if/then” decision trees that allow the system to react to a user’s immediate actions. If you buy a book about data warehousing from an online merchant, its system quickly flips out suggestions on other data warehousing titles that you should purchase. The system doesn’t look back at your entire buying history, or recognize that you just bought one of those suggested books last week, or discern that because of some complex combination of previous purchase and profile factors, you would be more likely to order a specific series of books. The book merchant’s “if/then” decision process – for speed’s sake – bases customization and suggestions on the immediate transaction, not on the longer history of previous ones.

Note the stark contrast between this “reactionary” cross-sell capability – quite similar to that which you might find in a call center environment (e.g., if a blue jacket is ordered, also suggest blue pants) – and the more complex approach we might take with outbound direct marketing messages. The former relies on the speed of a snap decision or reaction based on immediate data; the latter depends on the increased accuracy of a cross-sell projection based on a much larger data set and related analysis process.

– Application integration

Taking this a step further, wouldn’t the optimal solution let you have it both ways? In other words, to have the speed of the online bookstore’s cross-sell engine and the depth of data and related accuracy of a broad base of historic customer information?

Obviously. the Web could someday be driven by a depth of customer information. If a Web site could “take a look” at all of my previous interactions with the company, and also at projected value and cross-sell projections prior to coming up with a standard offer, wouldn’t this be a much more meaningful interaction for all involved?

While this last option appears to be the most logical, it is also the most difficult – and therefore seems to have few success stories thus far.

There are two challenges: The first is a lack of developed and proven systems to mine the historic data in real time to determine the optimal message and offer. The second is a lack of historic data to prove the potential return of a more targeted approach through systems integration.

Clearly, as e-commerce advances beyond a novelty, application integration will become more of a necessity.