How to read your Instagram Reels analytics

Instagram hands you about a dozen numbers per Reel. Three of them predict anything. Several of the rest are the same number wearing a different hat, and at least two go up simply because the post is old.

So this is a guide to the panel. What each metric is actually measured on, which ones Instagram itself has named, and the one comparison worth more than any raw figure on the screen. Every number below has a date and a source, because most of what circulates about Reels analytics has neither.

Start with the three Instagram named

On 22 January 2025, Adam Mosseri posted a breakdown of how ranking works. He named watch time, likes and sends as the top three signals, and told creators to pay attention to three specific things in their own insights: average watch time, likes per reach, and sends per reach. He added one nuance. Likes weigh slightly more for connected reach, meaning people who already follow you. Sends weigh slightly more for unconnected reach, meaning everyone else.

Read his instruction again, because most people skim past the important half. All three of the things he told you to watch are rates. Per reach. Not totals.

A total tells you how many people saw the post. A rate tells you what those people did once they got there. Only one of those is about the video. If you track totals, you will spend a year congratulating yourself for posts the feed happened to push and punishing yourself for posts it did not.

Views is not an audience

Meta's API defines views as the total number of times the media has been played on Instagram. Reach is the count of unique accounts that saw it at least once. Those are different questions and people constantly report the first as if it answered the second.

The naming got tidier in November 2024, when Meta folded Plays into Views across Facebook and Instagram. On Reels the calculation did not change, only the label. On Stories, photos and text posts, Views replaced Impressions and counts repeat views.

Which gives you one genuinely useful thing to do with a view count: divide it by reach. That ratio is roughly how much people replayed you. A Reel at 1.1x is being watched once. A Reel well above that got looped, which is a real signal about the video and one of the few things views knows that reach does not.

Beyond that, treat the view count as a description of distribution, not a verdict on your work. It is the number most contaminated by things you did not do.

Watch time comes in two flavours and only one is comparable

Meta's API exposes both. ig_reels_avg_watch_time is the average time spent playing the Reel. ig_reels_video_view_total_time is the total time it was played, replays included.

Total watch time is a volume metric. It rewards reach, so it mostly re-describes the view count. Average watch time is the craft metric, and it is the one Mosseri pointed at.

One trap. Average watch time scales with runtime. A 12 second Reel cannot beat a 45 second one on seconds watched, and it is not worse for failing to. Convert it to a percentage of your runtime, and compare it only against your own posts of similar length. Otherwise you will conclude that your short videos are weak when they are just short.

Skip rate answers exactly one question

On 22 August 2025 the @creators account announced a retention chart inside Reels insights, alongside skip rate. This is the most useful thing Instagram has shipped for creators in years and almost nobody is using it.

Meta's API reference defines reels_skip_rate as the percentage of "views from people who skipped during the first 3 seconds", divided by initial views.

Three things follow from that definition, and all three matter.

• It is inverted from the language everyone uses. "Hold rate" advice is skip rate subtracted from 100. A 40% skip rate is a 60% hold.
• It covers three seconds and nothing else. Skip rate has no opinion about second 12. If someone tells you it measures your video, they have not read the definition.
• Nobody has published a credible benchmark, us included. The "under 40% is healthy" figures floating around trace to marketing blogs, not to Instagram. Instagram has published no threshold. Use the median across your own last ten Reels as your line and ignore everyone else's.

What skip rate is genuinely good for is telling two failures apart. High skip rate with a decent average watch time means you have a good video behind a bad opening. Low skip rate with watch time collapsing at second eight means the opening worked and the middle did not. In the second case, every hour you spend rewriting the hook is wasted, and without this number you would never have known which case you were in.

The retention chart is the only per-second data you get

Read its shape, not its endpoint. Every retention curve slopes down. That is not a finding.

What you are looking for is a cliff: a specific second where the line drops. A cliff is an edit you can name. A cut that lands wrong, a claim you undercut, the moment the payoff stops arriving. A steady slope with no cliff means nothing is broken and you are looking at normal decay.

Do this on your best post as well as your worst. On the good one you are looking for how flat you managed to keep it and what you did there.

Sends, saves and the rest of the engagement column

Sends are shares in the API, and they are the one Mosseri singled out for reaching people who do not follow you. If growth is the job, sends per reach is your number. Not sends. Sends per reach.

Saves sit in Meta's total_interactions alongside likes, comments and shares, and Instagram has never named them among its top ranking signals in anything I could source. That does not make saves useless. It makes them a usefulness signal: people file away things they intend to come back to. Report it as evidence the post was useful. Do not report it as a growth metric.

Comments are worth reading and not worth charting. The count is a function of reach. The content of them is where the next three video ideas are.

Follows, and a warning if you use a third-party dashboard

The Instagram app shows follows on a post. Meta's public metric list for Reels does not include follows or profile visits, which are listed for feed posts and Stories instead. So if your reporting tool shows follows per Reel, ask where it came from before you put it in a client deck.

The same list does include reels_skip_rate, so a tool with the right permissions can pull that one. Most still do not.

Either way, follows per Reel are small numbers. A good post moves them by a handful. Do not build a strategy on a metric where the difference between a great week and a bad one is four people.

What is noise

• Total likes, total comments, total views, read alone. All three scale with reach. They tell you how far the post travelled, which you already knew from reach.
• Follower count. It is an outcome. It is not an input to anything you can control this week.
• Posting time and hashtag count. I could find nothing from Instagram naming either as a ranking input. Neither appears in Mosseri's list of three.
• Any benchmark without a date and a source. Reels analytics attracts confident round numbers with no origin. Check the origin. It is usually one blog post citing itself.

The comparison that actually tells you something

Here is the part that changes how the panel reads.

Rank your posts by how far each beat your own median, not by raw views.

Your median controls for everything you cannot change: your audience size, your niche, your cadence, how the feed feels about you this month. A multiple of your own baseline is the closest thing you have to an isolated measurement of the video itself.

Three rules for doing it properly.

• Median, not mean. One breakout post drags an average upward and then makes every normal post look like a failure for the next two months.
• Age-match. Views accumulate. Compare a post at day seven against your other posts at day seven, not against a Reel that has been collecting views since March.
• Do it per metric. Median views, median skip rate, median sends per reach. A post can be ordinary on views and a clear outlier on sends, and that post is teaching you something specific.

An example of how differently that reads. Say your median Reel does 4,000 views. Post A does 11,000, which is 2.75x. Post B does 6,000, but its skip rate is well under your median and its sends per reach is double your usual. Sorted by views, A wins and you go make more of A. Sorted properly, B is the one that changed behaviour, and A may just have been picked up by the feed. Only one of those two findings survives contact with next week.

A fifteen minute weekly pass

This is a spreadsheet job, and it stays a spreadsheet job for a long time.

• Pull your last eight Reels. Columns: views, reach, average watch time, skip rate, sends.
• Add two computed columns. Views divided by reach, and sends divided by reach.
• Take the median of each column. That row is your baseline, and it is the only benchmark you should trust.
• Flag anything at 1.5x median views or better. The 1.5 is arbitrary. Pick a line, write it down, keep it for a quarter so the numbers stay comparable.
• Open the retention chart for the best and the worst. Name the first three seconds. Name where the curve breaks. Two sentences each.
• Write one thing to repeat and one to stop. Then go film. An analysis that does not change what you ship is a hobby.

Where a tool earns its place

Before paying for anything, it is worth being clear about how far the free panel gets you. The pass above works. It stops working when you are running four accounts, or when the honest answer to "why did this one land" needs someone to actually rewatch the video frame by frame, which nobody has time to do every Monday.

That is the gap Silma fills. It reads every frame and hears every word of a short-form video, then returns a score, a verdict and a specific list of fixes. It ranks posts by how far each beat that account's own average, which is the same sorting rule as above, applied without the spreadsheet. The outlier feed does that to other people's posts, so you see genuine over-performers instead of big accounts having a normal Tuesday, and a pre-check reads your draft before you post it and tells you where it will lose people.

What it does not do: generate, edit, schedule, or post anything. It is the brain, not the hands. If you want a tool that presses publish, this is the wrong one.

FAQ

What is a good skip rate on Instagram Reels?
Nobody credible has published one, and Instagram has not. The thresholds circulating come from marketing blogs. Take the median skip rate across your last ten Reels and use that as your line.

What is the difference between views and reach?
Views counts plays, repeats included. Reach counts unique accounts. Views divided by reach is roughly your replay ratio, and it is the only thing a view count knows that reach does not.

Why do old Reels keep gaining views?
Views accumulate for as long as the post is being distributed. That is exactly why you compare posts at the same age rather than against whatever is sitting at the top of your profile.

Average watch time or total watch time?
Average, expressed as a percentage of your runtime, compared only against your own posts of similar length. Total watch time mostly re-describes reach.

Do saves help a Reel get more reach?
Instagram has not said so. Mosseri named watch time, likes and sends. Saves are counted in total interactions, and they are a good sign the post was useful, which is a different thing.

Can a third-party tool show me skip rate?
reels_skip_rate is in Meta's API, so yes, with the right permissions. Follows per Reel is not in the Reels metric list, so be suspicious of any dashboard that reports it.

Sources

• Adam Mosseri, Instagram, 22 January 2025. Watch time, likes and sends as the top three ranking signals; average watch time, likes per reach and sends per reach as what to watch in insights; likes weigh slightly more for connected reach, sends for unconnected. instagram.com/p/DFFyRp-pINJ
• @creators, Threads, 22 August 2025. Retention chart announced inside Reels insights, alongside skip rate. threads.com/@creators
• Meta, Instagram Platform API reference, media insights. Definitions of reels_skip_rate, views, reach, ig_reels_avg_watch_time, ig_reels_video_view_total_time and total_interactions; the Reels metric list, which excludes follows and profile visits. Accessed 8 September 2026. developers.facebook.com
• Meta for Creators, 14 November 2024. Views becomes the single distribution metric across Facebook and Instagram; on Reels the calculation is unchanged, on Stories and photo or text posts Views replaces Impressions and counts repeat views. creators.facebook.com