Why haven’t we solved event recommendations yet?


It’s Friday night. You come home after a hard week of work, sink into your sofa, and open up Netflix. Here’s what you see:

Granted, there’s a good chance your name isn’t “Guest,” but you get the picture.

Smart!

Netflix presents you with a list of movies and TV shows it thinks you might like. While Netflix can certainly get things wrong, more often than not you’ll find something on that list that tickles your fancy.

Question: Shouldn’t it be just as easy to find cool events to attend? To open an app that shows handpicked events based on your interests and past behaviour?

It should. Yet the event industry has been struggling to get this right for close to a decade.

You see, events are quite a different beast compared to movies or Amazon products. To understand why, let’s look at four specific challenges with event recommendations.

1. “Cold start”

Recommendation systems like those of Amazon and Netflix depend on huge amounts of data to provide suggestions.

When a new product comes up for sale, there’s virtually no information available about how it’s rated or perceived by people. Similarly, when someone signs up to Netflix for the first time, the system knows next to nothing about the person’s preferences.

As such, the system doesn’t have enough data to come up with a relevant recommendation for that product or user.

This is known as the “cold start” problem, aka “new user / new item problem.”

The cold start issue isn’t unique to event recommendations. However, it does come up far more often in the event space.

First, there’s a larger proportion of new users here; most people only sign up to get tickets to a particular event (e.g. to see a celebrity they already like). They’re unlikely to interact with the site after that, so the recommendation system has precious little info to go on.

Second, events are almost always “new,” since any information about ratings and preferences collected about an event isn’t directly valid once the event is over.

This relates to the next issue…

2. Events are one-off

Say you’ve read a book and loved it. You go to Amazon, give the book five stars, and write a gushing review. Others can now consider your opinion when deciding whether to buy this book.

Not so with events.

Once you attend an event, it’s over. You can tell others it was the best day of your life, but they can’t go back in time to experience that same event based on your recommendation. Time machines are hard to come by these days.

Events have a very short shelf life. This adds an extra layer of complexity when trying to find similarities between events and recommend relevant ones.

3. Events are location-bound

If Netflix thinks you might like a movie, it’ll safely recommend it knowing that you can watch the movie straight away, from the comfort of your living room, wherever you are in the world. (Unless you’re one of us poor souls who live in Denmark and only get access to about 30% of the US Netflix library.)

With events, your location matters. There might be an event that’s perfect for you happening somewhere tomorrow, but if it requires two flights and a three-hour kayak ride to get there, you probably won’t jump at the chance.

Sure, some of us might travel halfway across the world to see our favourite comedian perform, but such cases are exceptions rather than the rule. We might also see VR solutions that let you “join” any event right from your home. Eventually.

For now, event recommendation systems have to factor location into their algorithms when prioritising the list of suggested events.

4. Event recommendations are time-sensitive

Once you find a scalp massager you like or a book you want to read, you don’t have to act on it straight away. You can add them to your wishlist and buy / watch them later on.

Events happen at a specific time. You need to know about the event well in advance in order to make up your mind, free your schedule, and get the tickets.

So not only does an event recommendation have to be relevant and geographically close, it should happen early enough for you to act on it.


As you can see, recommending events to users can get rather tricky.

You’ll find apps and websites that offer curated lists of event recommendations. Some of these do try to personalise suggestions based on various factors. But nobody has yet succeeded at getting anywhere near the level of relevance that Amazon and Netflix have achieved with their systems.

So are event recommendations a lost cause?

Not at all! We believe it’s possible to improve the accuracy of event recommendations and get people to attend more events. In fact, that’s our very mission at Billetto.

How can this be done, exactly? One step at a time.


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.