The first rule of improv is always say, "Yes, and..." (though not literally). When I started taking improv classes, that was drilled in to my brain. Now that I'm in my third level, my teacher uses a new phrase on repeat: "If this is true, what else is true?" It's a way to heighten the scene.
I find myself hearing Keely's voice in my head at other times too. Like when I'm mapping content to a user journey. In improv, you don't know what you partner is going to say. If you create websites or products or apps, you don't always know what journey a user will take through your interface. Sure, you can make guesses, but you really need to create a framework to allow any journey to happen spontaneously and with meaning.
Recommendations
Amazon is famous for its Recommendations, as is Spotify. Frankly, I'm more inclined to trust Spotify's recommendations these days than Amazon's. You see, my kids sometimes use my accounts for their own shopping or music listening. I can tell who has been browsing Amazon when I login and see Big Nate books in the "New for you" list. And surely Amazon doesn't think the only books I'm interested in are old books about dystopian futures because of the last 4 books I ordered. What about the 20 before those? It seems that recency is the overriding factor in Amazon's recommendation algorithm. That's not super helpful.
But Spotify's algorithm uses more of a "If this is true, what else is true?" method for recommendations. If, in the past week, I've listened to 20 hours of Americana, indie, and classic alternative music but 2 hours of AC/DC and Journey, my Discover Weekly list isn't suddenly filled with 70s and 80s rock anthems. Because if Carrie listens to 20 hours of one type of music and only 2 hours of an entirely differently kind, it might be that someone else is listening, not Carrie. (More on Spotify's Discover Weekly Algorithm)
Sure not everyone has the deep engineering and scientific capacity that Spotify or Amazon have, but everyone can think like them. Artificial intelligence is getting more accessible every day. But if you aren't thinking about how to apply it or structuring your content in a way that is accessible to the machines, you're already behind.
Related content
Similar to recommendations, many systems try to make connections between one piece of content and others. All too often, there isn't much thought given to how to make these connections. At worst, every blog post has the same three blog posts in the "You might also like" list in the right column. (Pro tip: If recency is the only way you can connect content, don't bother.) At best there is a serious personalization algorithm at work. Most of us are somewhere between those extremes.
I used the "If this is true, what else is true?" line of thought on a recent project. We wanted to connect the construction projects in a meaningful way, and we wanted to limit it to three. At first, we planned to use a combination of category, location, and completion date to determine which additional project would show in the "Related Projects" area of a project detail page.
Instead of stopping there and trying to figure out the weighting of data for the algorithm, we started to think about what the primary users of the site would really be interested in. Iteratively, we trimmed it back to to one factor: location. Since this is a big public area, it is more likely that if someone is interested in a project in the southwest corner of the area, they probably don't care about the same category of project in the northeast corner. So we eliminated that input.
And they probably don't care so much about other projects that are going to be or were finished at about the same time. Eliminate completion date. The users are likely to want to know what else is happening or recently completed in their corner of the world. We ended up with "Nearby Projects" with a simple algorithm matching the location taxonomy of the project being displayed and sorted by completion date, in reverse chronological order (projects with the latest completion date first).
Always ask questions
These are just two of the ways asking questions – and one simple one in particular – can lead to more valuable and meaningful content connections. Do not make assumptions.
Content strategists and information architects: Don't make the designer or developer make stuff up. Have these conversations with stakeholders or clients as soon as you are thinking about connecting content. It's a business decision, not an implementation decision.
Designers: Don't make stuff up when you're moving pixels around. Ask someone who can answer specifically how these pieces of content are related. Labels matter, because words matter.
Developers: Don't guess when programming the algorithm. Ask someone who can answer specifically how these pieces of content should relate to each other. Also ask what fields to use for display and filtering and in what order to sort them. Tell the decision makers when there is a problem or something odd about the results.