Content Annotation, Taxonomy and Search Findability A conversation with Lynette Ledoux, Customer Success and Search Specialist, SearchUnify

It is a personalized experience. My kids freaked out when we launched the DJ and it said, “Hey Lynette, I know that you love rock and roll and traditional country from the fifties and sixties, so I've got one for you.” They were like, “Whoa. How does he know your name? How does he know you like that?” Well, because “he” has access to consistently labeled data and access to my data. Because I have a subscription, DJ can also see just the history of songs that I play apart from my playlist. What are playlists except folders and what's inside? The folders are songs, but the songs are digital files. Files have fields. The fields are artist title and genre. These fields are consistently labeled, which is exactly why when users go searching for different songs, they get a hit.

And also exactly why when you select a song to play the title and artist, at least oftentimes the genre displays song after song when you're playing it, this is what GenAI relies on to do what it does to find content, to deliver personalized results and to make accurate predictions. So when I launch the DJ, it looks at my history, it looks at all my playlists, and it looks at whatever other pool of content it has access to and makes selections according to whatever parameters the algorithm tells it to. The DJ will also tell you, “Hey, I'm going to change it up a little bit to kind of broaden your horizons, but if ever you don't like my selection, then you can, click the DJ button.” This is a data point that feeds the algorithm. If the DJ goes in a direction that I don't like and I click that button, that song will be removed from the pool of songs that the DJ can choose from.

Marydee Ojala: I get your point about personalization and about consistency and that those are important. And indeed, I totally agree. I guess my question is whose responsibility is that you have a personal relationship with your Spotify, but when we're talking about enterprise search, it's a much broader set of content and a much broader user base.

Lynette Ledoux: And a lot more specific to the use cases of the business. The magic of the Spotify DJ is facilitated by the fact that the employees of Spotify probably didn't have to go in and manually label all of these files. These songs have been digitized for decades. Multiple vendors use them and stream them. If songs happen to be missing labels or labels are incorrect, then there's ample opportunity to fix them.

That's not the case with enterprise knowledge. The companies are creating content specific to their products. The teams aren't necessarily collaborating, which produces an inconsistent experience, the total opposite of the Spotify DJ experience. Teams that are doing their due diligence and labeling content with, say, the product of an enterprise, maybe doing so with the most up to date product name. Or, they may be using a legacy product name and just never got around to changing it to the updated product name. Or they may be using an acronym. Some teams may be consistently using content types. Others may not. Some teams may be really trying to categorize issue types or sections of the customer journey that will be most applicable for the usage of the product. That's great. But if they're not coordinating, probably they're using different vocabulary to describe the same thing.