Building a consumer behavioral model using ML tools to optimize conversion paths in the ecommerce industry
One of the areas where the proper use of data can significantly improve the results of companies is the e-commerce segment. A few years ago, the issue related to generating traffic on an e-commerce website was analyzed regardless of the actions taken by the user on the website. The integration of these has been noticeable for several years by activities of entities specializing in the so-called “Customer Journey Optimization.” The new approach is holistic. It not only improves the results by matching resources to each user more effectively but allows for better and faster identification of changes taking place on the market (e.g., reasons for a drop in sales or the appearance of false traffic generated by bots). The process of optimizing the customer conversion path at an individual level is one of the greatest challenges facing the e-commerce industry. By using advanced analytics based on personalized decisions, we answer four basic questions necessary to optimize the conversion path:
Despite the dynamic development of technology, making many decisions in this process is still performed by domain experts. Human intuition is naturally biased, so marketers often use the method of carrying out the so-called A/B tests to evaluate individual elements of the customer journey. However, the presented method has many disadvantages, including in a specific A/B test. And therefore, we check the impact of one specific (expertly-selected) feature. But we ignore the possibility of the interdependence of many features on each other (e.g. the common influence of the background and the words used in the message), and through a manual process, we limit ourselves to specifically selected features and segments. The project’s challenge is to address this problem.
Enzode is a Polish e-commerce and Big Data company.
The project concerns the creation of the world’s first system to automatically manage the path of user-to-customer conversion in the field of electronic advertising.
The system, apart from a high degree of automation, will be characterized by the fact that it will cover the entire conversion path, i.e., from advertising, through the user’s behavior on the e-commerce website, to actions after leaving the website.
The applied approach will also allow for deep personalization of the message depending on the user’s profile. The effects of algorithmically-made decisions will be better than decisions made by domain experts.
Project in the testing phase. The algorithm already independently manages the
division of paths for parts of the campaign. The results are as good as obtained by humans or even better.