Man and woman shopping on mobile devices.
Machine learning hasn't yet delivered on all the hype, but it has improved specific areas along the customer journey PHOTO: rawpixel

Machine learning has come a long way in the last couple of decades, but it hasn’t come far as people sometimes imagine. Rather than AI taking over processes completely, it is helping to increase the efficiency of human-led processes. 

To quote noted IT author and professor Tom Davenport, “Augmentation means starting with what minds and machines do individually today and figuring out how that work could be deepened rather than diminished by a collaboration between the two.”

Still, some envision that AI will eventually take over completely, but for now machine learning can make processes, especially customer journeys, more efficient and successful.

How Machine Learning Improves Customer Experience

Machine learning can produce a more accurate and efficient taxonomy, product data presentation and product relationships for a buyer to choose from when shopping online. In the case of complicated technical products, a primary customer concern is having confidence in product selection. When many similar products could potentially to part of an engineering design, the need for each piece of that design to meet specifications is essential to project success. Machine learning can help this process in several ways.

Machine Learning and Classification Taxonomy

The first use is in creating the classification taxonomy, which allows each product to be classified to a terminal category (i.e., the lowest level sub-category). Next, clustering automatically groups similar products to suggest new subcategories in the taxonomy, making it even easier to find items.  Machine learning can be possibly used to build categories to aid in schema design (the attributes assigned to each terminal category).

Machine-learning-aided software can help pull attributes from existing data sources and recommend attributes to be included in the taxonomy. Companies can make their product data more consistent and efficient for customers. These category-specific attributes are then used as navigation filters on the website once the customer has selected a product category of interest. The capability to have these specification filters provides the technical buyer with the confidence needed to select a product for their engineering design and compare similar products.

Machine Learning and Product Classification

Second, machine learning is used in product classification. A new and more efficient classification taxonomy is now in place, but then comes the usual problem in classifying the existing items within the new terminal category. By classifying one or two products in each terminal node you can then use machine learning to recommend a category for the remaining products.

Conversely, it is possible to classify a set of recommended products into a defined category. This is a huge time-saver for someone who typically works in spreadsheets and databases and helps ensure that the product is categorized properly so it picks the correct set of category specific attributes. Because the products are already categorized into a group, and the group can be assigned all at once into a selected category thus allowing mass classification rather than singular instances.

The machine learning system is iterative and uses a virtuous cycle of machine and human aiding each other as more products are classified into a category. This allows the machine to produce better recommendations and classifications as it "learns" more about the products coming into the category. This same system that can help classify a product can help identify product relationships when you get to that a later stage.

You can’t have AI without IA (information architecture), which is the structure that organizes the data behind the machine. This adds to the cycle, making the machine more effective, which makes the user more effective. Aiding machine learning to find the data through clean taxonomy and ontology is the key to having a system that can be better trained and provide data more efficiently. This then help the user to be more effective and saves the customer time in their purchase, which is a huge benefit of using machine learning both on the back end and front end of customer experience.

Machine Learning Boosts Personalization

We are now seeing applications of machine learning that aid in creating a more personalized shopping experience. The system can track purchase and search history and recommend products or categories of interest as seen on Facebook and Google. More companies are starting to bring this capability into their ecommerce sites to aid customers in quickly finding a previously researched product or suggesting items they may like to purchase.

Chatbots have been pushed out to ecommerce sites over the last couple of years to aid the customer in finding the correct product, find current order status or ask general questions. Many customers prefer this method of communication over calling and talking to a person, so having this option appeals to a large set of people who want data faster and without the anxiety of calling a service center. This is another instance where the organization of data behind the machine intelligence is important. If the correct path to the data does not exist, the machine won’t be able to find it.

What’s Next for Machine Learning?

Apple now offers facial recognition for device security. Perhaps this technique will be deployed to further reach where customer history is automatically used for shopping preferences, not just on your device, but others as well. For example, I could research a car for purchase, enter the dealership knowing what model, make, year and blue book information, which reduces the anxiety that often comes with buying a new car. I could go to the MAC makeup counter and they would already know what color make-up would be a good match to my complexion.

The possibilities are impressive — this same scanning capability could someday measure foot size when going to a sports store for running shoes or remember your gaming history and automatically create characters geared to your preference. 

One thing we do know: businesses will continue to explore the limits of how machine learning improves the customer experience.