Most advice about the Instagram algorithm is a decade of folklore stacked on top of a handful of real, documented signals. This guide sticks to what Instagram has actually said, and what practitioners can verify from their own analytics.
There is no single algorithm
Feed, Reels, Stories, Explore and Search each rank separately. A post that performs well in Reels may barely register in the feed, because the surfaces are answering different questions. Reels asks what will hold a stranger's attention? The feed asks what does this person want from accounts they already chose to follow?
That distinction explains most of the confusion. When someone says "my reach dropped", the useful question is always: which surface, and compared with what?
The signals that genuinely move reach
Instagram's own guidance names these repeatedly: watch time and completion, likes, shares (sends per reach), saves, relationship history✓.
Read that list carefully, because the order matters.
- Watch time and completion dominate on Reels. Getting someone to the end, and ideally round again, is worth more than any number of likes.
- Sends per reach — how often people share a post into DMs — has become the signal practitioners watch most closely. It is hard to fake and it strongly predicts distribution.
- Saves indicate the post has future value. Carousels and reference content earn these.
- Relationship history decides whether your followers see you at all. Accounts you interact with get priority, in both directions.
| Surface | What it optimises for | What earns distribution |
|---|---|---|
| Reels | Entertainment for strangers | Completion, rewatches, sends |
| Feed | Relevance to existing followers | Relationship, saves, comments |
| Stories | Closeness | Replies, taps forward vs exits |
| Explore | Discovery by interest | Early engagement velocity |
What does not work the way people think
Hashtags are not a distribution lever. They help Instagram categorise content and surface it in topic feeds, which is worth doing — Instagram allows 30 maximum; 3–5 focused tags recommended✓ — but stuffing thirty tags on a post has not been a reach strategy for years.
Shadowbanning, as usually described, is not a thing. What does exist is content being made ineligible for recommendation: if a post brushes the Recommendations Guidelines, Instagram may keep showing it to your followers while withholding it from non-followers. Reach halves and nothing tells you why. That is a policy outcome, not a punishment, and the Account Status screen will usually say so.
Posting time matters far less than people claim. The feed is ranked, not chronological. A good post surfaces hours later; a weak one does not get rescued by a 7pm slot.
Engagement pods hurt. Engagement from accounts with no genuine affinity teaches the system to show your work to people like them — which is to say, people who do not care.
What to do instead
- Design for completion, not for the first frame alone. A hook buys you two seconds; structure keeps them. Reels can run to 3 minutes (up to 20 minutes for some accounts via uploads)✓, but length should follow the idea rather than the cap.
- Make something worth sending. Ask plainly whether a real person would forward this to one friend. If not, reach will stay inside your existing audience.
- Post at the size the surface wants. Reels are 9:16 vertical, 1080×1920✓; feed images work hardest at 1:1, 4:5 or 3:4 (4:5 recommended for reach)✓.
- Reply to comments within the first hour. It compounds relationship signals at exactly the moment ranking is being decided.
- Read the retention graph, not the like count. The point where people leave tells you what to cut next time.
How to tell whether your reach problem is real
Before diagnosing an algorithm change, separate three things:
- Reach per follower — is distribution down, or has the audience grown while absolute reach stayed flat?
- Follower vs non-follower split — a fall in non-follower reach points at recommendation eligibility; a fall in follower reach points at relationship decay.
- Format mix — swapping from Reels to static posts changes reach for reasons that have nothing to do with a penalty.
Most "the algorithm hates me" weeks turn out to be one of those three.
What the first hour actually decides
Early performance matters, but not for the reason people assume. Instagram is not rewarding you for a fast start; it is sampling. A post goes to a small slice of your audience first, and how that slice behaves determines whether the sample widens.
This has two practical consequences. The first is that the people most likely to see a post early are the people who interact with you most — so your most loyal followers effectively vote on whether everyone else sees it. The second is that a post which does badly in its first sample rarely recovers, which is why deleting and reposting sometimes appears to work. It is not a trick; it just buys a fresh sample.
What it does not mean is that you should chase engagement in the first hour with bait. Comments from people with no genuine affinity teach the system the wrong thing about who your content is for.
Stories and the closeness signal
Stories rank on a different axis again: closeness. Instagram is estimating how strong the relationship is between two accounts, and that estimate is built from DMs, profile visits, story replies and repeat viewing.
Stories segments run to 60 seconds per card✓ each, which is short enough that the format punishes waffle and rewards a single clear idea per card. The metric worth watching is not views but replies and forward-taps: replies build closeness, rapid forward-taps erode it.
For a business this is the most underrated surface on the platform. Feed reach is contested; story reach among people who already like you is comparatively cheap, and it is where the actual conversations start.
A sensible measurement habit
Pick three numbers and look at them weekly rather than daily:
- Reach split between followers and non-followers, which tells you whether you have a discovery problem or a relationship problem.
- Sends per reach, which is the closest thing Instagram gives you to a quality score.
- Profile visits from a post, which is the step before anything commercial happens.
Daily checking mostly measures noise. A week is long enough for a post to finish distributing and short enough to change what you do next.
The honest summary
The system is trying to guess what each person will enjoy, using behaviour rather than intent. Every legitimate tactic is a version of the same thing: make something specific enough that a specific group genuinely wants it, and make it easy to finish and easy to pass on.
