A bottom-up approach to product recommendation.
Response to my column on using metadata for personalization was overwhelming. Clearly, it touched a nerve. We discussed how to automate product recommendations by looking at the metadata associated with viewed products, and then recommend items with similar metadata. This is a top-down approach to product recommendations.
To continue with metadata conversion, let's look at more ways to discover patterns and trends. Look at what's beneath the products and data on your site: the metadata layer. This bottom-up approach assembles various pieces of metadata into a matching product. When might that be useful?
Guided Selling or Gift Purchasing
Sometimes you don't know exactly what you want. Perhaps you want a birthday present for a friend or a product in an unfamiliar domain (such as a new car). You don't know the actual product you want to buy, but you know either what your needs are or what features the product must have. Metadata again comes to the rescue, this time by enabling a configurator to help with your purchase.
A customer walks into a car dealership. The salesperson greets her at the door and asks if she'd like some help. She's looking for a little car for weekend trips in the country. "For a whole family or just you?" Just her and a friend. "Do you need a lot of room to carry things, or do you want something small and sporty?" Small and sporty. "Is this just a summer car, or do you need a car that's good in the winter also?" Just a summer getaway car. Something sexy and fun to drive.
After this short exchange, the salesperson takes the customer over to the roadsters. He shows her the BMW Z4, the Porsche Boxster, and the Mini Cooper. He also shows her the Mazda Miata and a small Toyota Solara convertible (evidently, a diverse dealership). Depending on how much money she wants to spend and what features she wants (e.g., heated seats, GPS, etc.), she chooses one of these five cars and further customizes it.
In only three questions, the salesperson was able to narrow the customer's choice to five cars out of hundreds of possibilities.
Back to the Online World
The salesperson used pieces of information (metadata) to assemble, from the ground up, a picture of the car that fit the customer's needs. He then took that information and recommended a small product set. If you've ever bought a car online, you know the process is similar. The automated configurator asks you a series of questions aimed at narrowing your choice to just a few cars.
Lots of science lies behind the question order. Starting the conversation with, "Do you need automatic windows?" doesn't reduce the product pool by much. It's the wrong initial question, though necessary toward the end of decision process.
Although this may be obvious to a human, it isn't to a computer. How does the computer know when to ask the right questions?
Smart configurators don't follow a script and ask everyone exactly the same questions in exactly the same order. Instead, they dynamically look at the taxonomy and ontology of the product space and figure out which set of metadata attributes represents the biggest difference between the remaining products.
The goal is to ask as few questions as possible while eliminating the largest number of products with each question. Asking about automatic windows might remove a few products from a pool of 600 cars, but asking what size car you need may reduce the product space from 600 to 70. Smart configurators can figure that out and ask the fewest questions possible.
Just for Cars?
Using configurators to suggest products isn't restricted to cars. They work great for gift buying if you know the type of things your friend likes, such as funny movies, self-help books, or pretty decorative things to put around the house. We built a gift advisor that served this purpose for Barnesandnoble.com several years ago. Within 5-8 questions, the configurator narrowed a product space in the millions to about 10 or 15 recommendations.
Even More About Metadata?
I've only scratched the surface of metadata uses. Product metadata is just one of several types of metadata to have in your knowledge database.
Customer metadata is the next large topic that really digs into metadata's power. Once you understand how to create and use product metadata and customer metadata, the real magic takes place: finding ways to combine these two sets of metadata to create any number of front- and back-end analytics to help you understand customers and their needs.
Stay tuned for more!
As always, let me know your comments or questions (or if you find this discussion of metadata useful). Your comments drive future topics.
Until next time...
Jack Aaronson, CEO of The Aaronson Group and corporate lecturer, is a sought-after expert on enhanced user experiences, customer conversion, retention, and loyalty. If only a small percentage of people who arrive at your home page transact with your company (and even fewer return to transact again), Jack and his company can help. He also publishes a newsletter about multichannel marketing, personalization, user experience, and other related issues. He has keynoted most major marketing conferences around the world and regularly speaks at Shop.org and other major industry shows. You can learn more about Jack through his LinkedIn profile.
2015 Holiday Email Guide
The holidays are just around the corner. Download this whitepaper to find out how to create successful holiday email campaigns that drive engagement and revenue.
Three Ways to Make Your Big Data More Valuable
Big data holds a lot of promise for marketers, but are marketers ready to make the most of it to drive better business decisions and improve ROI? This study looks at the hidden challenges modern marketers face when trying to put big data to use.
December 2, 2015
1pm ET/ 10am PT
Wednesday, December 9, 2015
5pm HKT / 5am ET