Fortnite Discover is where players can find and enjoy a wide variety of experiences. As a creator and developer in Fortnite, it’s important to understand how islands are surfaced in Discover. Focusing on key metrics, utilizing best practices, and understanding the Discover UI will help you attract, engage, and satisfy players, ultimately leading to greater success for your experiences.
This documentation provides a comprehensive overview for creators looking to maximize their island's visibility and success within Discover. For more context on Discover’s vision and goals, check out the Discover Vision Blog.
As Epic improves Discover in support of this mission, changes will be updated here. Refer to this doc whenever you have questions on Discover.
Below, you can find a comprehensive overview of:
The key metrics used to identify successful islands and the Discover Score.
How the Sophistication Score impacts testing of both new and updated islands.
How these metrics influence which islands appear in Discover.
Key Metrics
Discover is driven by three key metrics: Attraction, Engagement, and Satisfaction.
Attraction: Converting impressions into sustained engagement.
Engagement: Keeping players engaging, and retaining within your island.
Satisfaction: Players reporting your island as fun and highly rated, by players favoriting and recommending your experience.
These all combine to form a single Discover Score for each island.
Attraction
Attraction measures how you can capture a player's attention and convince them to play your island.
Attraction is measured by:
Clicks per impression.
Minutes played per impression.
Percentage of players who leave in the first two, five, or ten minutes of a session.
To be counted as an impression, your island's thumbnail must be 100% visible on the screen and viewable for two seconds or more. If a player clicks your thumbnail, even if it's not fully visible or does not appear for the full two seconds, the impression will still count as an impression based on the click-through result.
Click-through rate (CTR) is the ratio of clicks over impressions, measured as a percentage for how many impressions resulted in a click. For example, if your island has 100 impressions and 5 clicks, that is a 5% CTR. CTR measures the effectiveness of your thumbnail and title at capturing player attention and getting players to click.
Minutes played per impression is the ratio between impressions and total minutes played. For example, if your experience has 1000 minutes played in the last 96 hours after receiving 100 impressions in the same period, your minutes per impression would be 10. This metric shows whether your impressions are converting to sustained engagement.
Island details, or how you describe your island, can also influence player clicks once a thumbnail has caught their attention. You can update the island name and description in Island Settings. The information you put in Island Settings shows up in Discover on an island's details screen. You can also change this info in the Creator Portal.
Island tags are another way to help categorize your island in Discover. These tags are one of the inputs used to identify the genre of the game (see Category / Genre Grouping Rows).
See Publishing from the Creator Portal for how to add these tags. Also see Games and Game Tags for more information on the tags themselves.
For more tips on how to increase attraction, see:
Creator Marketing Playbook for ways to create key marketing assets that will catch player interest and increase click-throughs.
Thumbnail Image Guidelines for best practices on images in Discover.
Project Analytics to analyze patterns for new and returning players.
A/B Thumbnail Testing to validate the effectiveness of different images.
Engagement
Engagement focuses on keeping players invested and interested in your island once they've entered, and keeping them coming back.
Engagement metrics described here represent how Discover calculates engagement, it does not describe the Engagement Payout formula. To learn more, see Engagement Payout.
Engagement is measured by:
Minutes per player (typically a 96-hour lookback)
D1 (Day 1) retention
D7 (Day 7) retention
Number of Unique Players over the last 96 hours
Social play time spent in social parties
Discover does not count engagement while a player is idle. If a player hasn’t registered an input for sixty seconds, they are considered idle and these minutes will not count toward your island's engagement metrics.
To improve engagement:
Use Project Analytics with a focus on session engagement, retention, and minutes per player.
Also use the Insights Dashboard to learn more about player activity while on your island.
Continually iterate on game design, features, and work with your community to generate player feedback.
Satisfaction
Satisfaction offers qualitative insights on player experience within your island. This includes factors like how much fun players had, how they rate your game out of 10, and if they favorite or recommend the island.
Satisfaction is measured by:
Endorsements: Players who Favorite or Recommend your island. (Players have the option of clicking Favorite and/or Recommend when exiting a game.)
Post Match Surveys: Players who rate your island on fun, satisfaction, quality, or difficulty after they've played.
Surveys have a 2% chance of displaying to a player after a game session (unless the player has opted out of surveys). A player cannot be surveyed more than once every 14 days.
There are many possible surveys a player might be served, and not all of them relate to Discover.
Discover Score
Attraction, engagement, and satisfaction metrics are combined to create a holistic Discover Score that currently influences For You and Recommendations. The following metrics are used in Discover Score:
Unique players per impression over 96 hours is used to evaluate how well your island converts impressions into sustained engagement, drawing on metrics from the most recent 96-hour period.
Minutes per player over 96 hours and unique players over 96 hours are used to evaluate how long players engage with your island, and how many of them return to play later.
Fun survey scores over 96 hours indicate player satisfaction.
Most of these metrics are aggregated over 96 hours to consider players who engage over multiple sessions and across multiple days. Looking back over 96 hours also naturally incorporates retention.
These metrics update hourly to ensure there is rotation and that new islands can qualify. Changes to your island that improve these metrics should result in a direct increase in For You impressions.
Epic uses a modified version of what is conventionally known as a z-score to calculate the Discover Score and Sophistication Score. A z-score tells you how many standard deviations a data point is from the population average, letting us compare Discover and Sophistication Score metrics across different scales or distributions.
New and Updated Content Testing
Discover makes an effort to test every new and updated island to determine their potential for success in Fortnite. A test is defined as receiving 50,000 impressions, ideally within a 20–60-minute period.
Discover algorithms analyze an island's Attraction, Engagement, and Satisfaction during the test window. Islands with particularly high Sophistication Scores may be tested multiple times in their first week post-launch.
We recommend iterating and updating your island post launch after receiving feedback from your community and reviewing analytics in Creator Portal.
Island Prioritization and Sophistication Score
New islands are prioritized for testing based on their Sophistication Scores. Updated islands are prioritized based on development effort over the last 30 days.
The Sophistication Score uses the same statistical approach as the Discover Score, but focuses on the content within your island and the development effort spent building and updating your island. It includes metrics that focus on the usage of devices, animation, sequencers, Verse code, textures, and many other features. As new features are added to UEFN, they may be integrated into future calculations.
Re-using Verse code, devices, and systems from pre-existing islands is a standard and acceptable development practice, and will not trigger copycat detection.
How For-You Row Recommendations Work
The recommendations system in Fortnite combines player level personalization with Discover Score to deliver specific content to players based on what they are most likely to be interested in.
To determine what players are interested in and what content to serve them, a number of models and algorithms are used. The foundation for these models is the Island Embedding framework: the representation of each island as a standardized numerical encoding, designed to provide a data-driven and holistic understanding of each island in the ecosystem.
These Island Embeddings are formed by aggregating data like titles, descriptions, tags, thumbnails, and devices used. They also incorporate in-game telemetry that includes but is not limited to:
Player eliminations and NPC eliminations
Builds per minute
Structural damage
Vehicle usage rates
Distance traveled
Props used
Camera types
Weapons spawned and used
Each of these types of input data are consolidated and encoded to provide an accurate representation of islands across the ecosystem. Once these representations are established and standardized across all islands, they can be used to provide recommendations based on similarity, sequencing, and other factors.
These island representations also help to understand what each individual player is interested in. By algorithmically analyzing each player’s interactions — things like play history and Favorited Islands and Favorited Creators — it is possible to pinpoint the areas in Fortnite a player is interested in. Once those interests are established for each player, it's possible to serve them content from every part of Fortnite they might want to see.
Examples of models that operate on top of this representation framework include:
Item-based models: Using solely most recently played islands, these models surface similar experiences that a player could enjoy.
User-based models: Machine learning-clustering techniques on a player's play history can identify several areas of the ecosystem where a player spends time and surface compelling content from each of those areas for that player.
Sequencing-based models: Using patterns of play from the ecosystem, these can predict the next best island to play based on the player's recent play history.
To generate the list of islands that can be recommended, we use the top 200 islands per Game Collection and Featured Collection by Discover Score descending. We also apply filters for >= 1000 unique users over 96 hours and <= 45% of time played as idle minutes. Note that these filters can vary slightly by platform.
We also use the Discover Score within the individual models generating recommendations as additional modifiers to which islands are returned. Islands with a higher Discover Score are more likely to be recommended compared to similar islands with a lower Discover Score.
As Fortnite experiences become more diverse and innovative, Recommendations are a way to help match players with content they're more likely to find engaging.
Reviewing the Discover Rows
The rows that players see in Discover and the order in which they are listed, can change based on the player's cohort (age rating, platform, region, personalization opt-in) and the Discover team is frequently experimenting with new rows and orderings to understand how it impacts overall player engagement.
The names on the rows may also vary. The tables below are intended to provide the basis for each row and the islands displayed.
🧠 Personalized and Recommendation Rows
These rows are personalized for players based on their play history, favorited islands, and favorited creators.
For You Based on Your Play History | This row showcases islands with the highest aggregated Discover Scores, based on a combination of attraction, engagement, and satisfaction metrics. The islands shown are then personalized based on a player’s recently played experiences, Favorited creators, and regional play history. Finally, filters are applied using a Sophistication Score to reduce low-effort and misleading content. |
New and Updated for You | This row contains recently updated islands that players have played multiple times in their lifetime play history. It also contains new islands from Favorited creators and from the player's frequently played creators. |
🧩 Category / Genre Grouping Rows
These are categories that filter for islands based on template and genre.
Game Collections (categories of games) Games Built Using Popular Collaborations | This row showcases islands developed from brand templates, such as LEGO® Templates and Fall Guys. Islands in these rows are filtered by the island template used, and sorted by various engagement metrics. |
Featured Collections | These rows showcase groupings of islands organized by category and genre. Discover identifies the genres of an island through an array of machine-learning models working together to assess island genre or category. These models all operate on various combinations of the same island metadata used in the Island Representation Framework:
This system also utilizes a human-in-the-loop workflow, allowing Discover operators to review and update categorizations as necessary. This acts as reinforcement learning for the models. An island can be assigned a maximum of three different genres and two different categories, which allows an island to bridge multiple genres and appear in each relevant area of Discover. As an island is updated, and as more players engage with the island, the machine-learning categorizations are re-run to ensure islands continuously appear in the right place. |
🧑🎨 Editorial Rows
These rows showcase islands created by Epic and islands selected by Epic staff.
By Epic | This row features islands developed or published by Epic Games. The content is editorially selected and not impacted by player metrics. No third-party content is included. |
Epic’s Picks | This row features highly-polished, innovative islands selected by Epic’s editorial team. These selections highlight standout gameplay or creative use of mechanics. Islands in this row are manually curated and subject to quality reviews before being featured. |
🔥 Performance-Based Rows
Each of these rows are driven by overall metrics, as described for each category.
Top Rated | This row prioritizes islands with the highest aggregated Satisfaction scores (Currently prioritizes Fun ratings). Only islands with >= 25% D1 (Day 1) retention and >= 20 player survey responses can qualify. Sophistication Score filters aim to remove low-effort or misleading experiences. Frequently, this row is regionalized and sorted by metrics per country. |
Most Engaging | This row prioritizes islands with the highest minutes per player over the past 96 hours. Islands must have >= 25% D1 (Day 1) retention and have reached at least 100 concurrent users (CCU) in the past 96 hours to qualify for this row. When this row places islands into the homebar, there is also a 5M impressions maximum. Sophistication Score filters aim to remove low-effort or misleading experiences. Frequently, this row is regionalized and sorted by metrics per country. |
Popular | This row prioritizes islands with the highest minutes per impression over the past 96 hours. Islands must have an avg. CCU of >= 2000 in the past 48 hours, and >= 15 minutes per player. When this row places islands into the homebar, there is also a 5M impressions maximum. |
People Love | This row prioritizes Variety islands in the top 10% of D1 (Day 1) retention, followed by minutes per player in the last 96 hours. Islands must have reached at least 1000 CCU in the past 96 hours to qualify. Only islands categorized as Variety can qualify. When this row places islands into the homebar, there is also a maximum limit of 5M impressions per 100 hours. |
Fan Favorites | This row prioritizes Battle Royale and Combat islands in the top 10% of D1 retention (Day 1), followed by minutes per player in the last 96 hours. Islands must have reached at least 1000 CCU in the past 96 hours to qualify. Only islands categorized as Battle Royale and Combat can qualify. When this row places islands into the homebar, there is also a maximum limit of 5M impressions. |
Great with Friends | This row prioritizes islands with the highest percentage of minutes in 4+ player parties. Islands must have a max CCU of >= 100 in the past 96 hours and >= .33 minutes per impression. |
⏱ Lifecycle Rows
Rows that display recently played or recently published islands.
Recently Played | This row displays a player’s most recently played islands. It includes all island types without ranking or filtering, unless the island has become private or delisted. The list updates dynamically based on the player's session history. Islands in recently played last for 90 days before no longer considered recent. |
New and Updated | These rows display newly published or recently updated islands to allow them to reenter Discover. Discover also applies filters to reduce spam or unchanged islands. |
New Updates This Week / New Experiences | These rows are additional spaces to display new and updated content that may have been missed in the New and Updated testing flow. Discover also applies filters to reduce spam or unchanged copies. |
🎮 Genre / Play Style Rows
All genre rows are sorted by concurrent users (CCU) descending with the largest number of concurrent players first. There are many game genres in the ecosystem, such as Variety, Combat, Horror, Tycoon, Fashion, Driving, Prop Hunt, and more! To see the full list of genres, check out Games and Game Tags.
How Search Works
Search is an important way for players to find specific islands and genres within Fortnite, and uses a hybrid approach combining two algorithms: keyword matching and semantic search. Keyword matching returns islands that contain text that exactly matches the user’s query. Semantic search returns islands that are semantically similar to a user’s query.
This hybrid approach allows players to find content using generic terminology while also supporting very specific searches. For example:
| Search Algorithm Type | Search Term | Result |
|---|---|---|
Semantic | space | Islands set on the moon |
Semantic | scary tycoons | Horror-themed tycoon islands, Zombie tycoons |
Keyword | ffa | Free-for-all islands |
Keyword | Reload | Reload |
Once results are compiled from a search, they are algorithmically sorted based on features like:
Semantic similarity to the search term used
Keyword match (if present)
Concurrent users (CCU) at that moment, coupled with other variables
Confirmed recent search count
When search results appear, they are sorted to show the most relevant results to the player’s query. This is based on what players are most likely to be looking for and current trends in the ecosystem.
The search algorithm is continuously improved by integrating more data and variables on island metadata, performance, and player preferences. By continuing to iterate on the algorithm, we will be able to develop a more pointed and dynamic search experience based on current player behavior. We are also A/B testing different search algorithms to make sure we’re delivering the best search experience possible.
Closing Tips for Improving Your Island's Visibility
In this guide, we walked through how Discover works by explaining the metrics that contribute to the Discover Score and how the Sophistication Score affects your island’s prioritization and testing.
With this knowledge you can increase your island’s visibility by:
Focus on converting impressions into sustained play, and bringing players back.
Leverage your Creator Analytics by understanding how players will see, click, and engage with your island.
Iterate and update continuously to create a more engaging and fun experience for your players.
Epic is committed to providing the best experience for our creators. With a thorough understanding of how Discover works, combined with your island analytics, you can optimize your island for success.