At Snap, we want to empower you to learn about the world and have fun with other Snapchatters. On Discover and Spotlight, we help you do so by showing you content that is relevant to you, as well as breaking news or other hidden gems we think you may enjoy. This is not an easy task to do, that’s why we’ve developed a creative set of rules that determine what content to show you. Now we’ll walk you through the various steps we take when personalizing and ranking content on Discover and Spotlight!
1. Personalizing and Ranking Content on Discover and Spotlight
No two people are the same, so that’s why we use algorithms to personalize and rank content on Discover and Spotlight. We believe content is more relevant and entertaining when it’s personalized to your interests, and not someone else’s.
So what’s an algorithm? Put simply, an algorithm is a set of rules that if followed can solve a problem or lead to a solution. For example, a recipe for apple pie is an algorithm. Just follow the steps to get to a delicious result.
Of course, our algorithms are a little more complex than a recipe for apple pie. We use different algorithms that talk to each other, which improve the more you use our services. For example, if you watch a lot of sports content, and indicate that you don’t like hair and makeup tips, the algorithm will prioritize sports and skip those makeup tips. But, we won’t just show you sports content. We think diversity of content is important to avoid echo chambers.
2. Moderating Content
Our Community Guidelines outline specific rules for content that is prohibited on Discover and Spotlight. We want Snapchat to be a safe and positive experience for everyone so we prohibit content that, for example, contains any sexually explicit content, violence or dangerous behavior, facilitates illegal activity, or promotes false information.
We do our best to determine whether content is of sufficient quality to surface to the Snapchat community by automatic content moderation that is set to detect harm, as well as human review by one of our trained team members.
3. Understanding Content & User Preferences
To understand content and user preferences we have to do a couple of things. First, we categorize all content submitted to Discover and Spotlight by labeling the content—for example “sports,” “makeup,” or “dogs.” Both humans and computers help with this process. Our team looks at Discover and Spotlight content and labels it based on what they see. So for example, if a bunch of puppies are chasing a ball in a Spotlight submission, we may tag it “puppies playing with ball outside.” We also label content based on keywords that are used. So if you submit content to Spotlight that contains “my cute puppy,” we may pick up that phrase and categorize the content as “puppy.”
Once we’ve labeled the content, we move to the next step, which is understanding if you actually like the content you see. There are a number of ways we try to figure out what content you’ll enjoy. For example, if you subscribe to a Snapchatter, share a Spotlight with your Friends, or favorite it, we’ll use those as indicators that you enjoy the content. On the other hand, if you skip, report or hide content, we’ll use those as indicators that you don’t enjoy the content. Other indicators we use to see what you like are how long you watch a piece of content and whether you completed it.
4. Making Content Age and Language Appropriate
We want to make sure that the content you see is appropriate for your age and the language you speak. To do this, we look at your country, your language, the age you provided when you created an account, and the age we think you may be, based on your friends and your activity. We use this information to group you with similar users in your age group. This way we can identify content that is appropriate for you, in the language that you speak, and is localized to your location.
5. Ranking Content
Now that we’ve labeled the content and figured out what content you’ll likely enjoy, we can make sure you see that content in the right order: we call this "ranking."
We organize or rank content for Discover and Spotlight based on a combination of your preferences and our understanding of the content. This is done by predicting the probabilities of various actions you might take when seeing a given item of content, such as favoriting or sharing it, and the expected amount of time you’ll spend viewing it.
To make predictions about your actions, we use a variety of indicators. Because everyone is unique, exactly how much any given indicator will affect these predictions will vary between users. For example, past engagement with other content will (of course) be unimportant for a new user. That said, for Spotlight, the indicators that, roughly speaking, tend to have the highest importance in our ranking system include:
- Your country, the language that you speak, your age and gender so that we can show you local content that is appropriate and tailored to your demographic
- Content or creators we think you like because you have been viewing similar content, tapped "favorite," subscribed, shared or commented
- Content or creators we think you enjoy by looking at the Story’s content category (e.g. this story is about "puppies") and Story metadata such as how long you viewed a story, who the creator is and other information such as posted location, language and hashtags
- Content we think you don’t like because you have chosen to hide or report it
We look at other indicators as well, although they tend to affect the ranking of items less. This includes:
- Your preferred content categories like "puppies" or "sport" based on your content or engagement with content over time
- Information about the creator whose content you’re watching such as the creator’s country, language and the creator’s content categories (e.g this creator likes to post about “basketball” and “beauty products”) - this tells us that you may be interested in similar creators!
- We may show you content from a creator whose content you haven’t watched before because we know based on statistics (looking at the creator’s country, language age and gender) that the creator is typically liked by other users who share a similar demographic as you
- We may show you content from a creator whose content you haven’t watched before because you have been subscribed to similar creators
Lastly, we consider:
- Your device type, network type and region so that we can show you local content suited for your device!
- Your friend’s engagement with Stories - if your friend likes a Story we think it’s a good chance you will too!
- We may show you content from a creator whose content you haven’t watched before because we know (based on overall engagement with the creator’s content how other users have engaged with the creator) that the creator is typically liked by other users who share a similar demographic as you.
The predictions are then used to assign a score to each item of content. The items are then ranked according to these scores, with further adjustments and limits so a diverse range of content is shown.
By using this content ranking, we hope that the algorithm succeeds in showing you the content you like the most. If that’s not happening, you can help the algorithm by watching content you enjoy for a longer time, subscribing to creators you love, favoriting content, and hiding and skipping the content you don’t like.
The Importance of Diverse Content
We believe it’s very important that you see diverse content. So if you’ve been watching a lot of basketball videos, we may mix it up with the content you see a bit, even if we think that you or Snapchatters who are similar to you are most interested in basketball. So, you may see some soccer highlights, breaking news, or other hidden gems we think you may enjoy.
We believe this helps you to discover new interests, and it prevents you from entering an “echo chamber” of repeated, similar content, without differing viewpoints or experiences.
European Union Content Personalization Settings
If you live in the European Union, you have the option to disable personalized public content recommendations. In Discover and Spotlight, you can disable personalized content by either tapping on '...' then 'Why am I seeing this content?' which will take you to Settings, or you can go directly to Settings and 'European Union Controls.' When you disable personalization, public content will be recommended to you with basic data only, such as the language you have set on your phone, your age, and your country. You will still see content, but it will be more random and less relevant to your interests. Note, you may experience some delay before the content experience changes. If you have disabled personalization but wish to enable it again, you can do so either by tapping on the favorite icon (♥️) in Discover and Spotlight and then tapping 'Enable' in the 'Show More Personalized Content?' screen or by going to Settings in 'European Union Controls.'