You are getting prospective customers to your Web site with some smart, targeted advertising. But is all this traffic generating the revenue that you expect? On the Web, you don't have a sales team guiding prospects towards the products likely to interest them most. No one to whisper in the consumer's ear, “Wow, this dress looks really good on you,” or “Did you hear so and so's latest CD? It's so great.” On the Web, if you don't grab your visitor's attention fast, you will lose them.
Low conversion on the Web is becoming a cliché. So how do you go about improving your visitor-to-buyer ratios? Smart marketers are turning to targeting to get the right facts about who their visitors are, to deliver targeted content on their site and to turn “tire clickers” into buyers.
Targeting delivers the right content to the right person on the right page. E-commerce sites or content providers use targeting as part of their advertising strategy to identify their best customers and place banners on sites proven to attract such visitors. But while many companies use targeted advertising to drive traffic to their site, most fail to extend their e-targeting efforts within their Web site. Some use collaborative filtering techniques to cross sell and up sell visitors. Collaborative filtering analyzes the sequence of a customer's behavior in comparison to other customers who have exhibited similar purchase sequences. It enables Web sites to display a selection of related products to these customers. Collaborative filtering, however, is more of a recommendation tool than a precise predictive solution.
To convert more buyers into customers and improve returns, companies must apply the facts they gather about their visitors to predict the type of content each visitor wants to see. An online music store, for instance, can use targeting to predict the type of music that will most appeal to certain visitors. When those visitors land on its home page, the site can serve up promotions for artists that meet their tastes. As a result, these visitors are more likely to click through and spend more time on the site and, ultimately, buy more.
Targeting is based on predictive techniques deeply rooted in traditional direct marketing: profiling and modeling. These techniques have been adapted to meet the specific requirements of the Web.
Profiling is based on the premise that the behavior of a visitor can be traced back to some key demographic or site-specific purchase driver. On the Web, click-stream data provides a rich source of information about the behavior of site visitors and can be used to indicate future online behavior. Click-stream and log file analyses can tell you where visitors are arriving from, what they do while they are on your site and where and when they leave. Such information is limited in helping marketers understand the critical facts about who is buying and why.
In i-marketing, as in direct marketing, point-of-sale behavior is more useful when complemented with available demographic and lifestyle behavior. The most widely used sources of such information are offline consumer database compilers. For years, these companies have thrived on helping marketers put the right offer in front of the right consumer. And while the combination of online behavior data and offline demographic data has recently raised serious privacy considerations, we shouldn't forget that the Web has also brought in new opportunities for privacy never before considered in the offline world. On the Web, profiling can easily be conducted anonymously without posing a threat to the privacy of visitors.
While profiling helps marketers set the facts right, modeling is the science of predicting from those facts what visitors will do when presented with a specific offer. It automates the process of identifying consumer segments and determining the most appropriate Web content or promotion for each. Models establish the relationship between different demographic and behavior attributes to create larger segments that share the same characteristics.
The models then score the probability that a consumer in that segment will respond to a specific offer. For instance, a model will score the probability that working women, ages 19 to 29, single and who drive a foreign sports car are interested in opera. Using modeling scores, a site will then be able to provide the content most likely to be of interest to a visitor who falls within that segment.
The Web adds a whole new dimension to traditional modeling. In direct marketing, models age quickly. Data is valid for a limited period of time, and new promotions require new models. On the Web, new statistical techniques help adjust predictions in real time to reflect changes in consumer preferences, enabling marketers to keep pace with the rapidly changing needs of visitors.
If an artist's new release shows wider appeal than previous titles in the genre, for example, predictions can automatically adjust to identify the wider appeal of a particular promotion. With predictive scores adjusting automatically, new promotions can be quickly added without sacrificing the accuracy of the models or raising the cost of a campaign. With adaptive, or dynamic, modeling, marketers gain maximum marketing flexibility and are guaranteed predictions that reflect current activity on their site.
The Web is proving a highly effective medium for e-targeting. Very soon, we will begin to see the promise of true one-to-one marketing. But to reach that goal and generate higher returns on the Web, e-targeting must go beyond advertising. It must be used to deliver relevant content within a site and become the cornerstone of integrated marketing strategies.
