How does the Instagram algorithm work?
Instagram is the fourth social networking platform with the most users. Many of us had a question about the algorithm. “How does the Instagram algorithm work?” Through a real video he posted Adam Mosseri (CEO of Instagram) In June 2021, we found answers to our questions about the logic of this algorithm. Masuri explained why different algorithms were used instead of the chronological order of the homepage or stories as follows;
Looking at the data in 2016, they noticed that 70% or more of the content on Instagram could not be viewed. Also, almost half of his close friends’ posts get ignored because of the homepage rankings. For these reasons, they began to use “arrangement”. They’ve tried to determine the “interest level” of the content they’re likely to see by turning users’ behavior on the app into data. In this way, they aimed to increase the time they spent on the app by showing them the posts they were most likely to be interested in.
Another important reason to include these algorithms is to eliminate the “time-wasting” feeling people have. Because users use their time more effectively by seeing their close friends and loved ones and accessing the content they are interested in more easily.
Also read: The best website to increase real Instagram followers
Here’s how the Instagram algorithm works:
Home and story ranking
In order to rank posts, the first step is to look at stories you haven’t seen yet and interest them. To understand this level of interest, the data about the user is looked at, people who have sent stories and posts that have liked the user in the past. With the help of this data, a “score” is obtained that shows the user’s interest in these posts.
When calculating this score, questions such as whether this post is a video or a photo and what is its content appear first. Secondly, information about the user who shared this content is discussed. For example, the person who shared the post is your friend, how often they share it, etc. Thirdly, information about the user, that is, the user’s behavior while using the application, is taken into account. For example, data such as whether your likes are mostly videos or photos is used. The final factor used in calculating this score is the interaction between the user and the person who submitted the post. For example, do you often comment on the post owner’s photos, do you like his posts?
Given this data, they ensure that they increase the user experience in the application and highlight the people they care about on their homepages. Since human psychology is complex and not fully known, convergence is often used in these calculations. With this data, an attempt is made to predict the user’s movements.
The top 5 criteria in making these expectations are:
- time spent on the post
- The user’s probability of liking the post
- likely to get stuck
- likely to be preserved
- Likely to click profile
Instagram Discover is the part that works independently of the people we follow and is what matters most to us. Again, to find these topics and posts of interest, Instagram applies a filtering method called Collaborative Filtering. In this method, the posts liked by the user in the past are checked. With the help of the data of other users who liked the same post, similar content that the user might like is presented in the Discover section.
4 criteria to consider when selecting publications to be recommended in the Discover section,
- Data about the post (likes, comments, topics, etc.)
- Your previous interaction with the account that owns the post
- Own user activity (too many recipe videos)
- The frequency of the post owner
Based on this information, the number of likes, likes, and shares to be published will be estimated. In this way, an idea is obtained of the user’s interest in that post and discovery is organized accordingly.
As in the Discover section, the posts you come across in Reels are shown according to the topics that interest you, rather than the accounts you follow.
The tag is taken into account for the order of the video to be shown to the user in Reels,
- User activity (what kind of videos do you watch, do you like them)
- Interaction between the user and the owner of the post (whether he watched the video before, interacted with his previous posts)
- Details about Rails (when shared, number of video minutes, etc.)
- Data of the person who shared the rail
Given the time users comment on the post, save it, and spend time posting it, they will come across the app, and the content to be released also changes. Although Adam Mosseri has mentioned that the algorithms are not 100% efficient and we may encounter errors from time to time, with this rating system, the posts that catch the attention of the users are highlighted. In this way, it aims for users to spend more time in the app.
Looking at the data, we can easily see that as a result of introducing Reels and supporting Discover with an algorithm, Adam Mosseri’s target average usage time has increased. Although the use of algorithms is well implemented in Reels and Discover, it can cause significant oversight because relevance rating in stories and posts cannot be perfectly generated. This leads to negative comments and criticisms from some users.
Also read: How to market via Instagram