The post is up. Two hours later you open your phone, look at the numbers, and try to work out whether it worked. No number on the screen answers that, so you compare it to the last post, reach a verdict based on how you feel, and forget it by tomorrow.
The numbers are not the problem. Asking after publishing is the problem. A post with no metric attached before it goes out cannot be judged afterwards, because whatever came back, you will find a reading of it you like.
The question is asked too late
Say I publish something whose whole purpose is that people save it to come back to, and it gets 12 likes and 40 saves. That is a good post. Say I publish something whose purpose is to start a conversation, and it gets exactly the same numbers. That is a failed post. Identical numbers.
So success is not a property of the post. It is a judgement of the post against something you said beforehand. Without that, all you are doing is looking at numbers and picking the comfortable interpretation.
This is not an admin point. It is the difference between learning something from four posts a month, and spending two years publishing while feeling like you never understood what works.
What each number means, by the official definition
Before you pick a metric you need to know what each one counts. I am not taking this from an article. I am taking it from Meta's developer documentation, because that is the only place the definitions are written precisely.
From the Instagram media insights reference:
- Reach: the number of unique users who saw the post at least once. The docs mark it as estimated, not exact.
- Views: plays or views of the media. The docs label it a metric in development.
- Saved: the number of times a user saved the media.
- Shares: the number of shares.
- total_interactions: likes, saves, comments and shares, minus unlikes, unsaves and deleted comments.
That last one deserves a pause. Total interactions is a net figure, not a sum. If somebody likes it and then unlikes it, the number goes back. Which means the screenshot you took two hours after publishing can be higher than the real figure a day later, and that is not a platform bug.
The insights overview adds two more definitions worth having. Engagement at the media level is defined as the total number of likes and comments only, and Reach at the account level is a different thing from reach on a single post. So when somebody quotes you an engagement rate, ask which number is in the numerator before you react to it.
Four documented limits that change how you read it
These are in the documentation and almost never in the advice, and each one changes the verdict:
- Under 100 followers, some metrics are simply not available. The overview page states this plainly. If you are early, missing numbers are not a comment on your post.
- Data can be delayed by up to 48 hours. So a verdict at two hours is a verdict on incomplete data by definition.
- These figures are organic only. The docs state that comments, likes, views and total interactions count organic interactions and exclude ads. Publish something and then boost it, and you are looking at two separate halves.
- Impressions is deprecated. The docs state it is deprecated for media created after 2 July 2024, and remains available for older media. If your engagement rate formula divides by impressions, that formula has stopped working.
Stories have their own rules: metrics are available for 24 hours only, values under 5 return an error code rather than a number, and replies return 0 for users in Europe and Japan. A story showing zero replies does not mean nobody replied.
Pick one number before you publish
The habit is simple. Before publishing, write one line in a note. This post is for this; the number that will tell me it worked is this; the floor below which it failed is this.
Mapping the goal to the number looks like this:
| What the post is for | The number that matters | The number that misleads |
|---|---|---|
| Reaching new people | Reach, and the share who are not followers | Likes, which come from followers |
| People coming back to it | Saved | Views |
| Starting a conversation | Comments with actual words in them | Comment count, since emoji count |
| Someone sending it on | Shares | Reach |
| Getting a client | Messages or clicks | Everything above |
The third column is the useful part. The misleading number is not a wrong number. It answers a different question, and it is usually the larger and more flattering of the two.
On the broader limits of measurement, including the things that happen and never appear in any number, I wrote about that in how to measure marketing results.
Why saves and shares are worth more than likes
If I had to follow one number on social, I would take saves.
Not because an algorithm likes it. Because of what the action is: a like is a decision made in a second while scrolling, and a save is a decision that this will be needed again later. The second carries information about whether the content was valuable. The first carries information about whether it was pleasant to pass.
A share is stronger still and means something different again: a person put their own standing next to your words in front of people who know them. Things get shared for one of three reasons. They are useful enough that somebody needed them, they say something the sharer wanted to say, or they are funny. None of the three necessarily sells anything.
Which is the point worth stating: the post that gets shared most is not necessarily the post that brings work. I have seen that often enough to stop being surprised by it, and the writing side of it is in how to write a caption people reply to.
When a post worked and no number shows it
Two kinds of success never appear on a dashboard, and knowing them stops you from killing content that is working.
The first: a post that reached very few people, one of whom was exactly the right client and sent you a message. The numbers are dreadful and the result is excellent. This happens constantly in B2B and in higher-priced services.
The second: a post nobody engaged with, which people mention to you a month later. That is cumulative effect, measurement does not see it, and you will not know about it unless you hear it out loud from clients.
The inverse holds too. High numbers from an audience that is not your audience is noise rather than success. If that keeps happening, the question is larger than the post and belongs to platform choice, which I covered in do you need to be on every platform.
And if you want to build this habit properly and learn to read numbers rather than watch them, U Courses have a practical piece on marketing analytics for beginners and the five numbers that matter.
Honest limits
The definitions above are Instagram's. Facebook, TikTok and LinkedIn define the same words differently, and comparing an engagement rate across two platforms without knowing how each one calculates it is a comparison with no meaning. I read Instagram's documentation for this piece and nothing else.
There is no correct benchmark for reach or engagement. I am not going to tell you 5 per cent is good and 2 per cent is bad, because the figure moves with account size, sector and platform, and the numbers in circulation come from companies selling analytics tools. The only benchmark worth holding is your own average over your last ten posts.
And ten posts is not a sample. Noticing that your save-led posts beat your comment-led posts is a signal, not a finding. The decision deserves twenty or thirty posts against the same goal. Before that, you are collecting notes.
The simplest test beats any channel matrix: ask the last ten customers who paid you how they found you. The answers show which channels work, and which ones you pay for that nobody came from. That closes Part 5 of my free book Growth Engine, which places this subject inside a complete growth system: Part 5, Digital channels: how you reach them (in Arabic).