A big question for Web site designers continues to be this: what is the optimal number and placement of marketing call-outs on a Web page?
The self-service and forward-leaning quality of the Web has led to page designs that all too often serve myriad marketing goals. A single page can have any number of callouts in its layout, trying to catch the attention of a variety of consumers in different states of need.
It is very easy for this messaging complexity to overwhelm consumers, to the point where they are unable to focus on any of the messages put before them. Furthermore, it is possible for a consumer's attention to be hijacked by a lesser-value callout, which can mean of a loss of marketing gains as prime call-to-actions are eclipsed by less-significant ones.
Interactive marketers have traditionally dealt with these layout problems using information architecture, reporting tools and usability research to tune the design of their pages. My earlier work on Web page callout optimization has led me to another approach. We can solve these problems using dynamic messaging optimization embedded into our application servers.
More specifically, a genetic algorithm could be deployed on a site to optimize the placement and numbers of callouts within a page layout to grow, on an ongoing basis, a page's marketing gains.
This algorithm could receive performance feedback from the actions taken by the visitors to the site, which could then be transformed into automated decisions about how most gainfully to display the page for future visitors.
These are the advantages of the genetic approach:
- A genetic solution can continually seek optimal solutions to being gainful, which can adapt to changes in the marketing environment as they happen.
- It can be easily deployed in an application server environment such as J2EE, PHP, or .Net. This is especially true given its lightweight programming footprint and applicability to the widely used model view controller (MVC) pattern.
- Genetic algorithms tend to explore a wide variety of potential solutions, which can lead to solutions that would otherwise not be considered.
- The insights garnered from the use of these algorithms can be immediately applied to a site, even without human intervention. This works to limit the problematic latency in manual tuning methods created by a delay between observing a solution and enacting it.
- The genetic approach uses real-world feedback, which is not prone to the research biases inherent in laboratory usability studies.
Genetic Decision Making