Customer metadata is information about your customers. You already have a lot of it. Information such as name and address are the more standard pieces of metadata, but a much larger world of customer information exists beyond identifiers and demographics.
There’s a really interesting use for this information; the reverse of how it’s usually used. Customer metadata can be used to grow your knowledge about your own products.
User-Centric Site Information
We’ve talked about applying attributes (such as categorization) to products. We can also create a user taxonomy that contains lifestyle information and product affinities. This information can be gleaned in two ways: explicitly (asking users) and implicitly (observing site behavior). Either way, the data can be used in many different ways on a site, and in back-end analysis.
If your site’s designed in a user-centric manner, based on the largest client segments’ needs, you already collect useful user information. A site that features needs-based browsing is the first step.
For a pharmaceutical company, a good needs-based design might be based on all the medications someone takes, health problems that person has, and research areas that person is interested in. A retail site might be designed around a person’s lifestyle (a working mom or a frequent business traveler). A financial site might be designed around the person’s life stage (such as a newlywed).
Aside from creating a site tailored to these types of people, an entire taxonomy must be created and stored in the user’s profile. Tag each piece of product (or content) with its product-centric metadata (e.g., category or department) and user-centric data (e.g., audience segments the product is intended for). A very robust vision of your users emerges.
Mark users by their user-centric and their product metadata and you learn a lot about them. Hopefully, this much is obvious. What’s more interesting is the reverse: learning more about your own products.
Learning More About Your Products
It’s common to think of users as having profiles. Why can’t products have profiles? Not simple metadata profiles. I mean complex, dynamic profiles.
If your site has a robust set of user-centric metadata (including demographics and lifestyle data), you can learn a lot about the kinds of people who look at the various areas of your site. When a user looks at a product in the “Video Game” section of an online store, his profile is updated (adding implicit metadata that shows an interest in video games).
That user data (the anonymous stuff) can also be added to the product profile. Now, the marketer can look at products in her marketing database and understand what kinds of users are interested in which products.
A video game might be ascribed the traits of its users: males, aged 10-25, night shoppers. By exchanging information like this in the reverse direction, you actually grow a product database with an awareness of its target audiences based on actual site usage, not just marketing projections.
Consider it reverse append for products. Most people take product attributes and apply them to users. In this case, we’re taking user attributes and applying them to products.
Reverse Append in the World of Metadata
The next step is a bit more complicated and circuitous. If you’re appending user profiles with product metadata (based on browsing patterns, purchases, or stated interests) and user data to products (based on the same), the next step is to share product data in the user’s profile with other products that user looks at.
Doing so helps you understand what types of products are shopped for by the same types of people. Once you know that, a whole new world of automated cross-selling opens up.
Automated cross-selling and correlation is a science unto itself. Yet current technologies that look just at statistical purchases don’t consider user metadata in the process. They can tell you what products correlate with each other, but they can’t tell you the customer segments that would be interested in the correlation.
For example, these systems can tell you people who buy product A also buy product B. It also correlates product A with product C.
The reverse-append approach can tell you what type of users buy products A and B together and what type of users buy products A and C together. When you need to make a product recommendation, you can take the user into account rather than just base recommendations on flat product correlations. You can make smart recommendations based on product correlations, combined with knowledge of the current user. And that’s really cool.
Until next time…