IMPERVIR® / blog All articles →
IMPERVIR · BLOGAugust 31, 2026

9 min read

The Instagram feed: ten predictions and the first 10 seconds

Three predictions out of ten are computed by the card separately for the first post of the session. Plus the mechanics of the carousel, the reordering of its cards and what the platform calls informative content.

The Instagram feed: ten predictions and the first 10 seconds

Three predictions out of ten are about a single post — the very first one.

The following feed has a system card of its own, separate from recommendations. It has ten predictions, and three of them are computed separately for the first post you saw when you opened the app.

WHAT YOU TAKE AWAY

All ten predictions of the following feed, three separate predictions for the first post, the mechanics of the carousel and the fact that the platform may reorder its cards. Nine minutes of reading.

CONTENTS

01THE FIRST
PART 01INSTAGRAM FEED

Three predictions out of ten are about it

There is nothing like it in any other card.

First, about what this document is. Instagram has several feeds, and each has a system card of its own. Here we are talking about the following feed — the one that shows the people you follow.

Recommendations — that is, the showing of unfamiliar authors — live in a separate card with predictions of their own: a breakdown of the recommendations card. They must not be mixed.

The ten predictions the card computes for a post in the following feed.
The ten predictions the card computes for a post in the following feed.

Now the thing that makes this card worth reading. Three predictions out of ten are worded not about a post in general, but about the first post of the session.

How likely you are to spend more than ten seconds on the first post. How likely you are to skip over the first post in Feed. How likely the first post you see is to keep you spending time in your Feed during this session.

Meta Transparency Center · the “Instagram Feed” card

Ten seconds are named directly. And whether the first post keeps the person in the feed at all is assessed separately: that is, the platform predicts not only the reaction to the post, but the fate of the whole session after it.

WHAT THIS MEANS FOR THE AUTHOR

Landing first in someone's feed is a separate event with a separate assessment. A post that came first and did not hold attention affects more than itself. This is the only card where the platform describes such a role.


02CAROUSEL
PART 02INSTAGRAM FEED

The second prediction the neighbours don't have.

First on the card's list stands the prediction about the carousel: whether you will scroll through it in full. For a format many consider outdated, that is a notable place.

Its signals are more interesting. The count includes how many times you swipe to the next card, how many cards you view and how many there are in the carousel in total. And then comes a line that explains oddities authors have been noticing for years.

Whether the carousel was reordered in your Feed.

Meta Transparency Center · same place, in the signals for the carousel prediction

That is, the order of cards in a carousel may differ from the one you uploaded them in, and the platform counts the very fact of the reordering as a signal.

What the card computes for a carousel.
What the card computes for a carousel.

The practical conclusion here is a cautious one: since the order may change, relying on “the first picture hooks, the second explains” is risky. Every card has to work apart from its neighbours too.


03THE SKIP
PART 03INSTAGRAM FEED

The skip is counted in comparison with its neighbours

Not on its own.

Among the predictions there is the probability that the post will be skipped. And the main thing here is in the signal attached to it.

How likely you are to skip this post compared to other posts shown alongside it.

Meta Transparency Center · same place

The assessment is relative. A post competes not against an abstract standard, but against whatever ended up next to it in one particular person's particular feed.

Next to it in the signals stands the format as well: photo, video or carousel. That is, the platform compares the probability of a skip taking into account what the post is.

What the probability of a skip is compared with.
What the probability of a skip is compared with.

04INFORMATIVENESS
PART 04INSTAGRAM FEED

Informativeness decides the DM share

And the platform lists what it treats as informative.

A separate prediction is the probability that the post will be shared in a direct message. The signal for it is worded unusually specifically.

Whether the post is informative content (news, product reviews, tutorials, how-tos).

Meta Transparency Center · same place

The four examples in brackets are the answer to the question of what the platform treats as informative. A news item, a review, an educational material, an instruction. Not an opinion, not a mood, not aesthetics.

Recommendations have a similar prediction, only there the quality is assessed by a language model. Here the wording is simpler, but the list of examples gives more that is practical: how this is arranged in recommendations.

TWO CAVEATS FROM THE PLATFORM

The card opens with the words “Below are some of the significant predictions” — that is, the list is incomplete. And it ends with a warning that the models and input signals change often. The document was updated on the twenty-ninth of June two thousand twenty-six.

SOUNDS FAMILIAR?

If your followers have stopped seeing your posts, we'll look at the recent ones and tell you where the showing to your own audience is lost.

Go through my feed
05WHAT TO DO
PART 05INSTAGRAM FEED

What to do this week

Four steps straight from the card.

01Hold the first ten seconds
02Build a carousel out of self-contained cards
03Add what they call informative
04Compare format with format

First: look at your posts from the point of view of the first ten seconds. That number is named directly, and it is about the first post of the session — the one a person sees the moment they open the app.

Second: build a carousel so that every card works on its own. The order may change, and the pairing “the first hooks, the second explains” falls apart then.

Third: add to your posts what the platform itself calls informative. A review, an instruction, a breakdown. That is the direct route to the prediction about sharing in direct messages.

Fourth: compare posts within one format. The probability of a skip is counted taking format and neighbours into account, so comparing a carousel against a video is pointless.

THE CHECK TWO WEEKS LATER

Take five carousels and five single posts and compare them within their own groups. If the carousels consistently lose, look at the second card: it may be that it doesn't work without the first.


06WHAT NOT TO DO
PART 06INSTAGRAM FEED

What not to do

Four conclusions the card does not support.

Carrying conclusions about recommendations over to here
The following feed and recommendations have different cards, different predictions and different signals. These are two systems, not one with different settings.
Citing “a skip within two seconds”
There is no such wording in the feed's card — it comes from the Reels chaining card. Carrying signals between surfaces means making things up.
Building a carousel on the order of its cards
The order may be changed, and the platform counts the very fact of the reordering as a signal. Every card has to make sense on its own.
Treating the list of predictions as complete
The card opens with the words “some of the significant predictions” and warns that the models change often.

And the general extreme: thinking that the following feed decides everything. It is responsible for the showing to your own audience. The inflow of new people comes through other surfaces: Explore and recommendations.


PART 07INSTAGRAM FEED

What to do next

Two minutes for today.

Open your latest carousel and look at the second card separately, without the first. If it doesn't make sense on its own, a reordering costs you half the meaning.

If your posts have stopped reaching your followers, it is worth ruling out the simple causes first: how reach works on Instagram.

WANT TO SORT IT OUT FASTER?

We'll look at your posts and tell you what works for the showing to followers and what doesn't.

If you want your followers to see your posts — leave a request. We'll get in touch if there are slots available at the moment. There is usually a queue and a wait.

REQUEST
We'll go through your feed

We'll look at your recent posts and tell you where the showing to followers is lost and what to change first. Free, no obligations.


Frequently asked questions

How does the following feed on Instagram work?+
It has a separate Meta system card with ten predictions: whether you will like the post, whether you will skip it, whether you will scroll through the carousel in full, whether you will share it in a direct message, whether you will turn the sound on and how much time you will spend on the author's profile.
How does the following feed differ from recommendations?+
These are different systems with different cards. The following feed arranges the people you follow; recommendations show unfamiliar authors.
Is it true that the first post is counted specially?+
Yes. Three predictions out of ten concern it specifically: whether you will spend more than 10 seconds on it, whether you will skip it and whether it will keep you in the feed for the session.
Can Instagram reorder the cards of a carousel?+
Among the card's signals there is a line: “Whether the carousel was reordered in your Feed”. That is, a reordering is possible and is counted as a signal.
What does the platform treat as informative content?+
In the card this is spelled out in brackets: news, product reviews, educational materials, instructions. The signal belongs to the prediction about sharing in a direct message.
What is the probability of a skip compared with?+
With its neighbours. The wording is direct: how likely you are to skip this post compared to other posts shown alongside it.
How current is this information?+
The card was updated on 29 June 2026. The same place says that only some of the significant predictions are listed and that the models and signals change often.
← All articles Home →