Instagram has revealed one of its best kept secrets: how its algorithm works, the system that is responsible for classifying and displaying the publications, ads, stories and reels that appear on the social network. The platform offers this information to promote “transparency in its operation”, so that users are in control of their experience.
The first thing that Instagram has made clear is that the algorithm changes depending on each tab of the application. The same formula doesn’t apply to Posts, Stories, Explore, Reels or Search, instead each of these sections uses its own algorithm “tailored to how people use it”.
On the post board, Instagram considers recent posts to be posts just posted by people we follow, as well as posts from accounts we don’t follow yet and that the platform thinks might be of interest to us.
Read also Ada Sanuy
And how does Instagram know what interests us? Well, it takes into account a series of factors: “what and who has followed, liked or interacted with recently”. “We personalize the experience so that you try to achieve a balance between the content of the accounts that you follow and the content of the accounts that you do not follow but in which you might be interested,” says the social network on its blog.
It also considers other factors such as format, although there are other signals that it takes into account, in this order:
1 The activity. 2 Posts we like, share or save. 3 Information about the post: how many people liked it or when it was posted and what location was attached to it, if any. 4 Information about the person who posted. 5 The history of interaction with someone.
“At Feed, the five interactions we look most closely at are the likelihood of you spending a few seconds on a post, commenting on it, liking it, sharing it, and tapping on the profile picture. The more likely it is that you perform an action, and the more we will weigh that action, the higher up in Feed you will see the publication ”, specifies the social network.
The ranking does not change much in the case of Stories. “We start by defining all potential stories (excluding ads) shared by the accounts you follow, and we remove stories that violate our community guidelines,” they clarify. Other signals are also taken into account such as view history, i.e. how often you view stories from an account; Engagement history, which looks at how often you interact with that account’s stories; and closeness, referring to the probability that they will connect as friends or relatives.
In the case of Reels, it finds those that users might like, and then ranks them based on how interesting it thinks they might be to users. Also, to find the most entertaining Reels, survey users to determine if a Reel is worthwhile or entertaining, before collecting signals such as interaction history with the accounts they were posted from.
To give more prominence to the user, the company has also commented that it is testing a new system so that users can say what they want to see more, with buttons like ‘I’m interested’ and ‘I’m not interested’. They will also be able to proactively hide suggested posts with captions that contain words or phrases they don’t want to see.
