Instagram Insights explained, metric by metric
Most explanations of Instagram Insights are one person’s guess copied by the next person. The definitions below are quoted from Meta’s own Instagram Platform API reference, read on 8 September 2026. Where the API is vague, or flags a metric as estimated, that is said out loud rather than smoothed over.
This is the panel-wide glossary. If what you want is judging one Reel, how to read your Instagram Reels analytics goes deeper on that. If you are weighing whether you need anything beyond the free panel at all, that comparison is here.
Reach and views are not the same shape
Reach is the “number of unique Instagram users that have seen the reel at least once”. One person, counted once, no matter how often they watch.
Views is the “total number of times IG Media has been played on Instagram”. Same person, five watches, five views.
Two things follow. Views divided by reach is a replay ratio, not an audience number. And reach is labelled an estimated metric in Meta’s own docs, which matters if you are putting it in a report with a percentage sign after it. Quote reach as a round figure, not to three significant digits it does not have.
Views became the single distribution metric across Facebook and Instagram in November 2024, replacing Plays on video and Impressions elsewhere. If you have year-over-year charts that cross that date, the line changes meaning partway along.
Profile visits, and the trap in it
Profile visits is “the number of times your profile was visited”. Times, not people, so one curious person clicking through four times is four.
The part that catches people out: in Meta’s API, profile_visits is available for feed posts and stories, and not for Reels. The same is true of follows. So “how many follows did this Reel get me” is a question the platform does not answer at the individual Reel level, and any dashboard that shows you a confident per-Reel follows number is doing its own attribution. Ask it how.
Interactions is a net number
Total interactions is the “number of likes, saves, comments, and shares on the reel, minus the number of unlikes, unsaves, and deleted comments”.
The word doing the work is minus. This figure can fall after you look at it, without anything going wrong, because people unlike things and delete comments. If you screenshot it on the 1st and re-pull it on the 30th, a small drop is not a bug and not worth explaining to a client as one.
Watch time comes in two flavours
Average watch time is “the average amount of time spent playing the reel”. Total watch time is “the total amount of time the reel was played, including any time spent replaying”.
Total watch time rewards loops, which makes it flattering and nearly useless for comparing two videos of different lengths. Average watch time is the one to track over time, and it is the metric Adam Mosseri named alongside likes and sends when he described how ranking works in January 2025.
Skip rate answers one narrow question
Skip rate is “the percentage of views from people who skipped during the first 3 seconds of the reel”. It is flagged as both estimated and in development, so treat it as directional.
It is a hook metric and nothing else. A video can have an excellent skip rate and still lose everyone at 0:11. Low skip rate plus low average watch time is the most useful pairing in the whole panel: the opening worked and the middle did not.
Story metrics, including one that is zero on purpose
Navigation is “the total number of actions taken from your story”, made up of exited, forward, back and next story. Forward is someone tapping past you. Back is someone tapping to rewatch, which is the only one of the four that is unambiguously good.
Then the one worth knowing before you build a reporting template: story replies return 0 for stories created by users in Europe since 1 December 2020, and in Japan since 14 April 2021. That is in Meta’s docs. If you manage European accounts and your monthly report has a Story Replies row, it has been reporting a hard zero for years and nobody noticed because nobody expected replies to be high anyway. Take the row out.
What belongs in a monthly report
A report is read by someone who will not open Instagram to check it. That argues for few numbers, each of which survives a follow-up question.
- Reach, rounded, with the word estimated next to it once.
- Average watch time, tracked as a line over months, not a single figure.
- Sends and saves, which are the two engagement signals that mean intent rather than politeness.
- Your own median, and how many posts beat it. This is the only line that answers “is this getting better”, because it moves when your work changes and not when the algorithm has a good week.
- Follows, at account level, where the platform actually reports it.
Leave out impressions, because it no longer means what your older charts meant. Leave out total watch time, because loops inflate it. Leave out any per-Reel follows or profile-visits number, because the API does not provide one.
The metric none of these give you
Every number above describes one post in isolation. None of them tells you whether 8,400 views is good, because that depends entirely on what your account normally does. A post at 8,400 views on an account whose median is 1,200 is a genuine outlier worth studying. The same post on an account whose median is 40,000 is a bad week.
That ratio, the post against the account’s own median, is the comparison Instagram does not show you and the one worth building your reporting around. It is also what Discover ranks on, and what competitor analysis uses so a small account’s 4x post outranks a large account’s ordinary day.
Stop reading the panel one post at a time. Paste a Reel and Silma reads the frames, not just the numbers, and tells you why it beat the baseline. Lite is $9 a month. Check your first video →
FAQ
What does accounts reached mean on Instagram?+
Reach is the number of unique accounts that saw the post at least once, counted once each however many times they watched. Meta labels it an estimated metric, so it is a good headline number and a bad basis for a precise percentage. Views, by contrast, counts every play including repeats by the same person.
What does profile visits mean, and why is it missing on my Reels?+
Profile visits is the number of times your profile was visited, counting visits rather than people. It is missing on Reels because Meta’s API only exposes profile_visits for feed posts and stories. The same applies to follows. A tool that shows you a per-Reel follows figure is estimating it.
Why did my interactions number go down?+
Because it is a net figure. Meta defines total interactions as likes, saves, comments and shares minus unlikes, unsaves and deleted comments. It falls when people undo things, which they do. A small decline between two pulls is normal and not worth flagging in a report.
Why are my story replies always zero?+
If the story was created by a user in Europe, replies have returned 0 since 1 December 2020, and in Japan since 14 April 2021. This is documented behaviour in Meta’s API reference, not a bug in your account or your dashboard. Remove the row from European reporting templates.
Which Insights metrics actually predict anything?+
Average watch time, sends and saves are the ones worth tracking, and watch time is among the signals Mosseri named publicly in January 2025. Everything else is context. The strongest single measure is not in the panel at all: how far a post beat your own account’s median.
Sources
- Definitions of views, reach, profile_visits, follows, total_interactions, ig_reels_avg_watch_time, ig_reels_video_view_total_time, reels_skip_rate, shares, saved, replies and navigation, their availability by media type, the estimated and in-development flags, and the Europe and Japan story-replies behaviour: Meta, Instagram Platform API reference, media insights, read 8 September 2026.
- Views replacing Plays and Impressions across Facebook and Instagram: Meta for Creators, 14 November 2024.
- Watch time, likes and sends named as ranking signals: Adam Mosseri, Instagram, 22 January 2025.
- Retention chart and skip rate shipping in Reels insights: @creators, Threads, 22 August 2025.
Claims we could not source and therefore did not make: a benchmark for a good skip rate, a target average watch time by niche, and any per-Reel follows attribution. Found something out of date? Email support@silma.app.