Please welcome: Superusers!
They’re not the heroes we deserve, but they’re the heroes we need right now.
It’s…perhaps less exciting than it sounds. Still, it’s nothing to sneeze at.
Let us explain.
Back in the ancient, dark days of our event recommendations journey (two weeks ago), we treated all visitors the same way when calculating similarities between events.
But let’s face it: Not all visitors are created equal. Sorry, Kyle.
Some people use Billetto more frequently and attend more events than others. We’d like their behavior to have a higher weight when looking for event similarities.
That’s why our model now looks at so-called “superusers” who:
- Placed four or more orders in 2016.
- Those orders were not for their own events (if they’re organisers).
We then look at the events these superusers bought tickets for and — importantly — those they didn’t. This lets us better predict how similar events are.
Here’s how that works in practice:
Say we know that Mary bought tickets to a concert.
Those concert tickets were on sale for two weeks. We now look at what other events were on sale during that same time period. We pick 20 of these at random.
This tells us that:
a) The concert tickets would’ve been a good recommendation for Mary at that time.
b) The 20 other events would have been…less…good.
With us so far? Great! Let’s dig a bit deeper.
Using this knowledge, we can make a comparison between all the other events Mary attended in 2016 and the 20 she didn’t.
By cross-checking them in terms of description, location, timing, and so on (as described here) we can now:
- More confidently identify similarities between these events.
- Find out which of the factors (description, location, etc.) matter more and then refine our model accordingly.
So there you have it: They might not have the ability to fly or turn invisible, but superusers possess the power to incrementally improve event recommendation systems.
Where’s our blockbuster movie, Marvel?!
Watch this space.
Every Thursday, we’ll be posting about the promise and challenges of personalised event recommendations, along with Billetto’s current efforts and future plans.
Have some thoughts on or experience with event recommendations? We’d love to hear them. You can leave a comment or send an email with your thoughts to dagn@billetto.com. We’ll read it. Promise.
