Instagram analytics: the 7 metrics that actually matter
You open Insights, see four tabs, thirty numbers and a row of “+18.4%” arrows, and close it again. A week later your client asks “so how are we doing on Instagram?” and you say “growing, I think,” because no honest answer assembles itself out of that screen.
The problem isn’t too little data. It’s too much of it, and most of it describes the past instead of pointing at a decision. A metric earns its place only if there’s an action attached: you see the number, you know what to change next. Everything else is report decoration.
Below: where the numbers live, the seven metrics that change what you do next, three that lie to you, how to read Stories, and a 15-minute weekly ritual.
Where the numbers actually live
- Insights in the app — the fast layer. Requires a professional account (Business or Creator); switching is free and reversible. Ranges are 7, 30 and 90 days against the previous period. Audience demographics (cities, age, active hours) only unlock at 100 followers — below that the section is empty, and that’s not a bug.
- Per-post insights — under the post itself, where the useful numbers hide: reach split into followers and non-followers, profile visits, and follows attributed to that post.
- Meta Business Suite on desktop — Instagram and Facebook in one window, period comparison and export. The only free way to get a spreadsheet instead of a screenshot.
One habit that pays for itself immediately: export to a spreadsheet once a month. Instagram isn’t obliged to keep last year’s history for you, and “how did this compare to the same month last year?” always comes up.
Seven metrics that change decisions
1. Reach, split into followers and non-followers
Reach counts people, not screens. The value isn’t the total — it’s the split.
If 85–90% of reach comes from followers, your content never leaves your own base and growth will only come from ads. If non-followers are more than half of reach but follows stay low, the problem isn’t the content — it’s the profile: your bio and first nine tiles aren’t explaining why anyone should stay.
2. Views
In 2024 Instagram collapsed impressions, plays and views into a single metric — Views — and made it primary across every format, from feed to Reels to Stories. Replays count.
On its own it says little. The useful part is views divided by reach: 2,000 reach and 3,400 views on a Reel means people are rewatching — a ready-made brief for a follow-up video.
3. Sends — the growth metric
Sending a post into a DM is one human recommending something to another, and the algorithm reads it as a much stronger signal than any tap. A like costs half a second; a send means “look at this, we need to do this.”
The takeaway: stop asking for likes and make posts that are easy to forward — price breakdowns, mistake teardowns, pre-purchase checklists. A direct ask at the end of a carousel (“send this to whoever runs your Instagram”) measurably lifts the number: people rarely do what nobody asked them to do.
4. Saves
A save means “I’ll need this later”: prices, instructions, templates. If saves are high but sends are flat, the content is useful but not social — people use it, they don’t share it. Add a point of view or one arguable claim and the forwards appear.
5. Profile visits → follows
Every post contains a small funnel: seen → tapped through to the profile → followed. Track both steps.
A real example: 3,000 reach, 90 profile visits (3% — fine), 12 follows (13% of visitors — good). Lots of visits and almost no follows means the post works and the profile doesn’t: rewrite the bio and the pinned posts. Almost no visits on big reach means nothing in the post gave anyone a reason to learn more about you.
6. Reels retention
Reels insights include a retention graph and average watch time. Reading it is straightforward:
- A cliff in the first 3 seconds (more than a third of viewers gone) — the problem is the opening frame and the first sentence. Nothing else matters yet.
- A steady slide — the video is longer than its idea. Cut it.
- A bump in the middle — people are scrubbing back to a specific moment. The best signal in Reels analytics: you found what they came for, so make a whole video about it.
7. Net growth: follows minus unfollows
Unfollows aren’t failure — they’re a filter, and they never stop. The signal worth chasing is a spike on the day of a specific post. Find it and look honestly: harsh tone, hard sell, a topic that missed. One teardown like that beats a month of staring at the follower chart.
| Metric | What it tells you | What to do if it drops |
|---|---|---|
| Non-follower reach | Whether content escapes your base | More Reels and shareable topics |
| Views / reach | Whether videos get rewatched | Make a follow-up to the winner |
| Sends | Willingness to recommend you | ”Show a colleague” formats, direct ask |
| Saves | Practical usefulness | Specifics: numbers, steps, lists |
| Profile visits | Interest in you as an author | Give a reason to learn more |
| Follows per post | Whether bio and pins work | Rewrite bio and first tiles |
| Reels retention | Strength of the first seconds | Reshoot the opener, cut length |
Three metrics that lie to you
Likes. The cheapest action on the platform, and ranking stopped leaning on them long ago. A mood indicator, not a basis for decisions.
Total follower count. A 40,000-follower account with 900 reach is 39,000 people who never think about you. Watch reach relative to your base instead of the base itself — how to calculate that ratio honestly, and what counts as normal, is covered in our guide to engagement rate.
“Best time” from built-in insights. It shows when your audience is online, not when it reacts — a hypothesis for the first test, not a schedule. How to validate slots on your own data is in our piece on the best time to post.
Stories: three numbers instead of twenty
Stories bury you in per-slide reporting, but three things decide everything:
- Retention from first slide to last. A 10-slide series with 1,100 views on the first and 380 on the last is 35% — normal for a long story. Below 25% means you’re stretching four slides’ worth of material.
- Exits on a specific slide. One slide where people leave en masse is worth a week of guessing: almost always a wall of text or an abrupt jump into selling.
- Replies and sticker taps. Polls, quizzes and question stickers aren’t entertainment — they’re how you grow the share of your audience that keeps being shown your Stories.
Taps forward and back only mean something next to exits; alone they’re noise.
The 15-minute weekly ritual
- Log the week’s posts in a sheet: date, format, topic, reach, sends, saves, follows.
- Find the best and worst post by sends. Not likes — sends.
- Write one hypothesis about why it won: topic, format, opening frame, timing.
- Test it next week with one similar post, without rebuilding the whole strategy.
- Check unfollow spikes and tie them to specific posts.
- Compare month to month once a month, not week to week — weekly swings are noise.
Two months of this gives you something no external guide can: a list of formats that work for you specifically — the easiest raw material for next month’s content plan.
When the numbers can’t be trusted
- Under 500 followers. The difference between posts drowns in randomness — one share from a friend doubles your reach. Watch topics and comments, not percentages.
- The first 48 hours. Reels and carousels keep collecting views for weeks, and one viral post wrecks your averages. Report on the median, not the mean.
- Any post with ad spend behind it. The moment you boost, organic and paid reach blend and the post stops being comparable to the rest.
- Other people’s accounts. You can’t see their ad budget, geography or history. The only honest benchmark is you, one month ago.
If you run several accounts and don’t want to retype numbers every week, Repost21 pulls Instagram, YouTube and Threads analytics into one view, and the calendar suggests slots from that same data. Try it free — one account is enough to tell whether you need it.