As evidenced by the COVID-19 crisis, e-commerce has become an integral part of our everyday lives. This migration to the digital world has caused organizations to ask themselves several questions, mainly, how do we monetize our digital content, and how do we do it effectively?

E-Commerce and Machine Learning

One solution is to bring an artificial intelligence (AI) and machine learning (ML) driven shoppable media experience to the consumer. This type of solution enhances the customer’s shopping experience by converting static media content to a dynamic interactive and shoppable media experience that truly engages the customer and reduces buyer resistance and increases electronic cart size, and thus, sales.

As we all know, innovative concepts and technology enhancements vastly improve the consumers' shopping experience. When this happens, it can lead to greater levels of engagement and ultimately higher check-out cart totals. (Read: Preparations for the Creation of an Online Store.)

However, while innovation and technology offer numerous benefits, there can also be undesirable results. For instance, organizations may lack the personnel to properly identify and associate products within the media content, which can create a barrier to offering a shoppable media experience.

A critical component for the creation of an interactive shoppable media experience revolves around automation. The good news is that today, automation typically comes with some level of machine learning, and we are seeing more and more machine vision being utilized within the shopping experience itself.

Machine learning and automation have allowed companies to optimize the shoppable experience for the consumer as well as streamline business processes and enhance profitability for the organization. (Read: How BIg Data Can Drive Smart Customer Service.)

Potential Hurdles to a Shoppable Media Experience

Let’s take a moment to look at a potential hurdle that impacts the monetization of digital content and creates barriers to providing a high level of consumer engagement. One such hurdle that organizations face as they try to move toward an engaging shoppable media experience is how to identify the products quickly and easily within the content.

As we discussed earlier, automation can help. By harnessing the power of machine learning and machine vision, we can make the identification process for the shoppable media experience much smoother with limited or no human interaction.

The Experience

One way to make an experience seamless is to automate the process of selecting products that are related to the shoppable media content and then present the results to a human for review and final validation. This methodology provides a safety net against product selection errors, while reducing the overall level of human effort required. (Read: A Brief History of AI.)

Among the numerous benefits to organizations that utilize these machine learning/automation techniques is that it will reduce their staff’s product selection efforts by 1-2 hours a week.

Case Study

Imagine that you are a retailer in the beauty industry and your organization has an extensive collection of products for sale. You currently utilize brand based and influencer video content to showcase your products. One of the videos in your collection takes the consumer through a journey of an influencer’s daily makeup routine.

Through the power of machine learning, the first step in the process of utilizing your product catalog is to vectorize the item images. This is the process of converting image files into a format that is usable by machine learning algorithms.

The next step is to perform multi-label classification and build a taxonomy around your offering. By utilizing a multi-label classification as well as a robust taxonomy, the algorithms will help you identify the product traits most applicable to the media content. These traits for this beauty product example may include the following:

  • Primary Color.
  • Secondary Color.
  • Item Type (Blush, Lipstick, Foundation, Brushes, Nail Polish.)
  • Design (If the item has any patterns on the packaging for instance.)

The final stage as video content is produced is to run the videos through one or more convolutional neural networks utilizing the product vectorized images previously created.

By running the video through the neural network(s), the seller can find the nearest neighbors (i.e. the closest match evaluating multiple vector points) across the various labels, along with a probability score for each match. Weighing this score allows them to find the exact match or the closest match to the identified product.

Final Thoughts

This example only scratches the surface of what is possible with machine learning and artificial intelligence. There is a whole world of innovative opportunities in this space ahead for the marketplace. These concepts can be extended even further to bring in additional ancillary data such as analytics, product inventory data and data management platform (DMP) data to enable further personalization of the products that are offered to the consumer.