purplETHought #2 Twitter's "For You" Feed

How to rank your tweets higher after another update to the algorithm

In my last post, I talked about how the open-sourcing of the Twitter algorithm would lead to a more frequent need to change and adapt the algorithm. It happened again: Yesterday, Twitter released a new update on how the “For You” feed algorithm works.

Nota Bene: Elon Musk announced that starting April 15th, only verified accounts will be eligible to be in For You recommendations.

Decoding the For You feed

At heart, the algorithm works by calculating a score for each tweet that tries to predict its engagement. The formula can be broken down into three steps.

1. Predicting engagement probabilities: For each engagement type (favorite, retweet, reply, etc.), the algorithm predicts the likelihood that a user will engage in that specific way.

2. Calculating weighted engagement scores: Each engagement type has a weight assigned, as provided in the update. To find the weighted score for each engagement, the algorithm multiplies the probability of that engagement by its weight (I will break that part down in an example).

3. Sum the weighted engagement scores: Add up all the weighted engagement scores to get the total score for the tweet.

Understanding the formula

The three steps above are encompassed in this formula.

score = sum_i { (weight of engagement i) * (probability of engagement i) }

Let’s assume a tweet has the following predicted probabilities:

  • Probability of favorite: 0.3

  • Probability of retweet: 0.4

  • Probability of reply: 0.1

The weights are as follows (they aren’t assumptions; we know the data from the code)

  • Weight of favorite: 0.5

  • Weight of retweet: 1.0

  • Weight of reply: 13.5

The score for this tweet would be:

score = (0.5 * 0.3) + (1.0 * 0.4) + (13.5 * 0.1) = 1.95

The higher the score, the more likely the tweet will be ranked higher in the "For You" timeline. The algorithm considers various factors to predict engagement probabilities, such as the tweet's content, engagement history, and user relationships.

Leveraging the algorithm

What can you do to improve your ranking?

  1. Focus on engagements: Create content encouraging users to engage with your tweets. Retweets, replies, and favorites have the most weight in the ranking algorithm.

  2. Encourage profile visits: Craft a compelling Twitter bio and share content that prompts users to visit your profile, as the algorithm considers good profile clicks a positive signal. Use ChatGPT if you struggle with the bio.

  3. Use multimedia: Include videos in your tweets if you can. The algorithm gives a small weight to video_playback50, meaning that users who watch at least half of the video in a tweet contribute to a higher ranking.

  4. Foster quality conversation: The algorithm rewards tweets that generate "good clicks" Share content that drives discussions and encourages users to participate.

  5. You can build relationships: You can encourage interactions with your audience, as the algorithm places a high value on replies that the tweet author engages.

  6. Avoid offensiveness: Twitter severely punishes offensive content that gets reported.

I think 1, 4, and 5 stand out the most. They aim to reward genuine interaction between humans. I can’t repeat that often

enough: If someone engages with your content, don’t take that for granted.

That person sees a thousand content pieces daily and chooses to talk to you. This has meaning. Build on it.