Optimizing your Tweets to Maximize the Algorithm
April 1, 2023
Twitter recently opened sourced their ranking and tweet visibility algorithms and here are some take aways on how you can improve your tweets so that are more likely to be boosted by Twitter.
According to twitters blog, Twitter's recommendation algorithm aims to deliver the best and most relevant content to users by selecting top tweets for their home timeline's "For You" feed. The process consists of three main stages: candidate sourcing, ranking, and applying heuristics and filters. The Home Mixer, a custom Scala framework, serves as the software backbone connecting these components.
Candidate sourcing retrieves the best tweets from In-Network and Out-of-Network sources. In-Network tweets come from users you follow, with the Real Graph model predicting the likelihood of engagement between users. Out-of-Network sources, on the other hand, analyze the social graph and use embedding spaces like SimClusters to find relevant tweets from users you don't follow.
After gathering approximately 1,500 candidate tweets, the algorithm ranks them using a neural network trained on tweet interactions to optimize for positive engagement. Next, heuristics and filters are applied to create a balanced and diverse feed, implementing product features like visibility filtering, author diversity, and content balance.
Finally, the Home Mixer mixes tweets with other non-tweet content, such as ads and follow recommendations, before sending them to users' devices. The entire process takes place in under 1.5 seconds and runs about 5 billion times a day. It’s a very large and complex system but you can definitely take actionable steps to make sure your tweets are going as far as possible!
To get the most visibility on Twitter and take advantage of Twitter's recommendation algorithm, you should consider the following tips:
- Engage with others: Engage with users who share similar interests or are part of your target audience. This will increase the likelihood of your tweets being recommended to them through Twitter's Real Graph model.
- Post relevant content: Make sure your tweets are relevant to your audience's interests. The algorithm looks for content similarity, so posting tweets related to popular topics in your niche increases the chances of your tweets being recommended to others.
- Optimize tweet timing: Post your tweets when your target audience is most active. This will improve your chances of being included in the In-Network source recommendations.
- Encourage engagement: Craft tweets that encourage users to like, retweet, and reply. The algorithm ranks tweets based on their engagement scores, so more engagement will increase your tweet's visibility.
- Connect with influencers: Engage with influential users within your niche or community. The SimClusters model uses influential users to identify communities, so connecting with these users can increase the likelihood of your tweets being recommended to a broader audience.
- Use relevant hashtags and mentions: This can help increase your tweet's visibility in search results and the Explore tab, making it more likely for your tweets to be picked up by the recommendation algorithm.
- Diversify your content: Post a mix of text, images, videos, and links to keep your timeline fresh and interesting. This can help maintain a diverse feed, which is one of the goals of Twitter's recommendation algorithm.
By incorporating these strategies, you can optimize your tweets to gain more visibility on Twitter and get the most out of the platform's recommendation algorithm.
Since Twitter has recently opened sourced their code, here are some takeaways derived directly from the code and tweet rankings. The ranking process relies on various features, boosts, and penalties applied to different tweet attributes.
Twitter uses a class called
DefaultFacetScorer. The class is responsible for scoring and ranking facets of tweets based on various features, such as user reputation, favorites count, language, and tweet content flags (offensive or containing multiple hashtags). It also applies penalties based on certain criteria.
The class uses several parameters, weights, and penalties to compute the score for each tweet, and it increments the facet counts according to these scores. The scores are calculated by combining the different features, applying penalties if needed, and considering language preferences.
Tweets that are generally ranked/weighted higher:
- Tweets from accounts the user follows or has interacted with in the past
- Tweets with high engagement (favorites, retweets, replies)
- Tweets containing popular or relevant hashtags
- Tweets from verified or influential users
- Tweets that are part of a trending topic or conversation
- Tweets with rich media content (images, videos, links)
- Tweets from users who have a mutual follow relationship with the viewer
- Tweets that match the user's interests or preferred language
Tweets that are generally ranked lower:
- Tweets from accounts the user doesn't follow or has shown no interest in
- Tweets with low engagement (few or no favorites, retweets, replies)
- Tweets containing irrelevant or spammy hashtags
- Tweets from users with a low follower count or low influence
- Tweets that are unrelated to current trends or conversations
- Tweets with negative content or flagged for sensitive information
- Tweets that are deemed less relevant or interesting to the user
Finally favorites and retweets are rated the highest in terms of improving Tweet performance!
In conclusion, optimizing your tweets to maximize Twitter's recommendation algorithm is crucial for gaining visibility and engaging with your target audience. By understanding the inner workings of Twitter's algorithm, you can tailor your tweets to rank higher in users' feeds and increase your reach. Implement actionable steps such as engaging with others, posting relevant content, optimizing tweet timing, encouraging engagement, connecting with influencers, using relevant hashtags and mentions, and diversifying your content. Additionally, by analyzing the open-sourced code, you can gain insights into the factors that influence tweet ranking, and use this information to create better-performing tweets.