Reach, saves, and the numbers worth reading

Instagram Insights shows you around nine numbers and gives no indication which of them deserve your Tuesday morning. So most people pick the one that is easiest to understand, watch it every day, and end up worse informed than if they had looked at nothing.

Three numbers are worth a weekly look. A few more are worth a monthly one. One is worth checking roughly never, and it is the one almost everybody checks first. Before any of that is useful, though, you need to know what each figure actually counts, because two of them sound identical and are not.

What each of your numbers actually counts

Start here, with no interpretation. The middle column is the definition, the right column is the only thing the number can honestly be used for.

Two rows in that table are the ones that cause trouble, and it is worth slowing down on them.

Reach counts people, views count viewings

Meta's own Business Help Centre defines accounts reached as the number of unique accounts that have seen your content on screen at least once, and impressions as the total number of times your posts have been seen. It states the difference directly: reach is different from impressions, which may include multiple views by the same accounts.

So one person watching your reel four times is one reach and four impressions. Views behave like impressions rather than like reach: repeats count. This means a reel with 40,000 views and 9,000 reach and a reel with 40,000 views and 33,000 reach are two completely different events. The first was watched again and again by a smallish group, which is usually a good sign about the post. The second went out wide and was watched once each, which is usually a good sign about distribution.

There is a practical consequence. If you only ever read views, you cannot tell growth from repetition, and you will misread the single most important question a small account has, which is whether anything you make is getting outside the group that already follows you.

The three worth looking at every week

Weekly is the right cadence for exactly three numbers. Here they are, with the reason each one earns the slot.

1. 01Shares, or sends. In January 2025 Adam Mosseri told creators that the three signals mattering most for ranking are watch time, likes and sends, and that in your insights you should watch average watch time, likes per reach and sends per reach. Sending a post to a friend is the most expensive thing a viewer can do cheaply: it puts their own judgement on the line in a private conversation. Instagram treats that seriously, and so should you.
2. 02Saves. Instagram has never named saves as one of its top ranking signals, so treat this as an argument about usefulness rather than about distribution. Somebody saving your post is telling you the thing had value beyond the moment they saw it, and that is the closest a public number gets to somebody saying your work was worth keeping.
3. 03Reach. The only one of the three that answers whether you are still reaching beyond the people who already follow you. Watch it as a share of your follower count rather than as a raw number: reach of 40,000 means something very different at 3,000 followers and at 60,000.

Everything else on the list is a monthly or an occasional number. Profile visits and follows-from-a-post are worth reading once a month, together, because they answer a question weekly noise cannot: whether the posts that travel are also the posts that convince. Impressions are worth a look when a post behaves strangely and you want to know whether people watched it twice. Likes are worth almost nothing on their own and are worth a little as a ratio against reach, which is the form Mosseri named.

The follower count is the worst number to check daily

It is the number every new creator watches, and it is close to useless at a daily resolution. It moves for reasons unconnected to anything you did: someone deletes an account, a batch of inactive accounts gets removed, a reel from three weeks ago resurfaces in somebody's feed. Read at a day's resolution, it is almost entirely noise with a slow signal buried inside it.

Instagram's own head has said as much. In April 2024, as reported by Social Media Today, Adam Mosseri argued that follower counts are not a good indicator of an account's value, and he has spent years pushing creators towards reach instead. He runs the company that displays the number, which makes the statement unusually credible: nobody at Instagram benefits from you caring less about a figure the app prints in bold.

There is also a straightforward psychological cost. Checking a slow number quickly is the recipe for demoralisation, because at a daily resolution the number is flat or down about as often as it is up, no matter how well the account is doing. Look at it monthly and it becomes a real signal. Look at it every morning and it becomes a mood.

Compare a post against your own middle post

The most common way people read their numbers is to look up a benchmark and see whether they clear it. Here is the arithmetic behind the most quoted benchmark of all, so you can run it before deciding what it is worth. An engagement rate is normally the interactions on a post divided by your follower count. Likes plus comments, divided by followers, multiplied by a hundred.

Now the parts that number cannot see. It divides by followers while most of a reel's reach may have come from people who do not follow you, so a post that travelled widely can score badly on it. Most versions of the formula count only likes and comments, so saves and sends, the two things Instagram and your own audience care most about, are invisible to it. It says nothing at all about whether anybody followed you afterwards. And on a single post it is noisy enough that a difference of a percentage point means nothing.

Then there is the benchmark problem. Rival IQ's 2025 Social Media Industry Benchmark Report puts the median Instagram engagement rate by follower at 0.36% across all the industries it tracks, with higher education at the top on 2.1% and the top quarter of brands overall at 1.02%. That is a six-fold spread depending on whose accounts got measured.

A benchmark is the middle of somebody else's accounts. Yours has a different size, a different subject, a different country and a different posting rhythm, and any one of those moves the figure more than anything you will change this month. So use your own middle post instead. Take your last twenty posts, line up their reach from smallest to largest, and take the one in the middle. That is your median, and it is the only fair thing to hold a new post against.

Use the median rather than the average, because averages get dragged around by one unusual post. Two months ago a reel did four times your normal reach: it will sit in your average for a year, quietly telling you that every ordinary week is a bad one. The middle post is immune to that.

When a bad week is only a bad week

Some weeks are down and mean nothing. Knowing which ones is the difference between adjusting and thrashing, and the reason a single post is noisy is public knowledge rather than a mystery.

TechCrunch reported in April 2024 that Instagram ranks recommendations in stages: eligible content goes to a small audience first, the pieces that group engages with go to a slightly wider audience, and the best of those go wider again. Your post's fate is therefore decided by a small first group. Who is in that group, what they had just watched, what time it was where they live: none of that is about your post, and all of it moves the outcome.

• One post is one sample. The gap between your best and worst post in a normal month is usually larger than any change you made on purpose.
• Posting time shifts who is in that first audience, and the same post at 8am and 8pm meets two different rooms.
• A holiday, an election, a cricket final or a national news event moves everybody's numbers at once and has nothing to do with you.
• Format changes the baseline. A carousel and a reel are not comparable, and comparing them produces a decline that never happened.
• Instagram runs tests continuously. Some weeks you are in one.

So the honest rule is a run of three. One post below your median is noise. Three consecutive posts below your median, in the same format, is worth investigating. A whole month below your median is a real change and you should go looking for what it was. Anything shorter than three posts and you are reading tea leaves, and you will change something that was working.

The mirror of this matters just as much. A single post that goes unusually far is also mostly luck, and the mistake people make after a good one is bigger than the mistake they make after a bad one: they rebuild their whole plan around a sample of one, and then spend six weeks confused.

What our software reads, and what it says when it does not know

This is the section about the product. Everything above works with a spreadsheet and a Friday afternoon, which is the arrangement this journal makes with you.

Synclify reads an account against its own history rather than against a published band. It keeps your posts, holds the numbers Instagram stops holding, and states what it finds as sentences: your widest posts were saved less than your ordinary ones, this format has beaten your median four times running, you have not posted in the pillar that carries most of your saves since June.

proof: findings come in 20 kinds, and the label set is a total record over that list, so a finding without a label fails the build rather than shipping unlabelled (shared/finding-labels.ts)

Underneath that there are 15 views onto the same account data, from the raw dataset through to a 100 point scorecard, and you can ask for any of them in plain words. The scorecard exists because a single number is easier to act on than thirteen screens, and it is a hundred points rather than a grade so you can see which part of it moved.

proof: 15 views and a 100 point scorecard (server/agent/tools/intelligence.ts, server/intelligence/scorecard.ts)

Questions people ask about their numbers