The Rise of Relevance

Search is becoming more important every day. If you’re creating an app, that familiar search box is probably a key part of your UI. While a great search UI is important, the fact is, voice is becoming a growing interface. In fact, during the Google I/O keynote, the CEO of Google stated that 20% of searches are now voice searches. Whether someone uses a voice search, or a UI, they’re expecting the first result, to be the most relevant result.

Relevance has been on my mind a lot for sometime now. I created a training course on Azure Search and that process made me think very deeply about relevance. It’s easy to just return any old search result. But, it can be tricky to identify which results are the most relevant. For example, in the training course, I show developers and engineers how to build a search engine for beers. Now, if you were searching for a beer, what would you want to see? What would make a beer the most relevant?

  • Is it the availability of the beer relative to your location?
  • Is it the name of the beer?
  • Perhaps it’s the brewery the beer comes from?
  • Should you take into consideration personal preferences? For example, maybe you like beers with a wheat flavor or perhaps more bitter beers.
  • Maybe the age of the beer should be a factor?
  • How should unstructured information, like a tweet be factored in?

Each question is a legitimate question on its own. Now, think about going a step further, and wanting to designate the most relevant beer as a combination of several of these. Like a phoenix, the question starts to rise of which of these questions is the most relevant. After all, that’s the question that will need to have the most weight. Now think about what you need to answer to connect users with your data.

Figuring out those questions can be a challenge. Coming up with a strategy that connects someone’s questions with your business’s data and goals can also be demanding. If you don’t nail down those questions though, your users will be forced to filter and sort through data. As of today, users are used to having to do this because it’s common. With relevance in play though, this approach can start to take a back seat. In a graphical user experience, this just adds additional convenience. When interacting with a voice assistant though, returning the most relevant result becomes a necessity.

While the training course I created is about building a search engine, I believe this is a misnomer. In reality, I believe the real value is in creating a relevance engine.