Blog
Reach, impressions and views are three different numbers
They are not synonyms. Each one counts something different, the platforms draw the lines in different places, and comparing the wrong pair is how a good month looks flat.
If you have ever looked at two charts of the same week and seen two different stories, this is usually why. Reach, impressions and views count three different things, the platforms do not define them identically, and a dashboard that puts them side by side without saying which is which will let you draw a confident conclusion from the wrong column.
Here is what each one means, where the definitions diverge, and which comparisons are safe.
The three definitions, in one sentence each
Reach is the number of distinct accounts that saw the content at least once. One person seeing a post five times is one unit of reach.
Impressions are the number of times the content was displayed, whether or not anyone new saw it. That same person seeing the post five times is five impressions.
A view is a play or a qualifying display of the content itself, and it is the one with the most variation between platforms and formats — a video view has a duration threshold behind it, and an image "view" often does not mean the same thing at all.
The relationship holds in one direction and only one: impressions are greater than or equal to reach, always, on any platform, for any piece of content. If you ever see reach exceed impressions, something is wrong with the report, not with your account.
Why the same word means different things on different platforms
Because each platform defines its own metrics, and they change them.
The clearest example is Meta's own naming. Over the past few years Meta has retired several metric names outright and introduced replacements that count something adjacent but not identical — for instance moving publishers toward a "views" family in place of some older impression-style metrics on Facebook. When a metric is retired, a tool that was reading it has to pick a successor, and the successor's definition is rarely a perfect match.
That is not a scandal, it is just how platform APIs work. But it has a practical consequence worth taking seriously:
A number is only comparable to another number produced by the same definition. If the definition changed in March, your February and April figures are two different measurements wearing the same label.
This is the single most common cause of a "sudden drop" that nobody can explain. Before investigating a cliff in a chart, check whether the metric behind the chart was renamed, retired or redefined on that date. Very often the content did nothing at all.
Checked 2026-08-27. Metric names on both Facebook and Instagram change on Meta's schedule, not ours — treat any specific metric name you read anywhere, including here, as something to verify against the platform's current documentation.
The comparisons that are safe, and the ones that are not
| You want to know | Use | Not |
|---|---|---|
| How many people saw it | Reach | Impressions — repeat displays inflate it |
| How heavily it was shown | Impressions | Reach — it cannot rise above 1 per person |
| Whether people watched | Views, with the duration threshold stated | Impressions — a display is not a watch |
| How often the average viewer saw it | Impressions ÷ reach (frequency) | Either number alone |
| Whether this month beat last month | The same metric, from the same source, over equal-length periods | Two metrics that merely sound alike |
Two rules make almost all of this go away:
- Compare a metric only to itself. Reach to reach, views to views, over periods of equal length. Comparing this month's views to last month's impressions produces a number, and the number is meaningless.
- Compare within one source. The platform's own app, its API, and any third-party tool can all be right and still disagree, because of timezone boundaries, attribution windows and deduplication rules. Pick one source for a given question and stay in it.
What engagement rate has to do with it
Engagement rate is a ratio, so it changes meaning depending on which of these three numbers is in the denominator — and the denominator is frequently not stated.
- Engagement ÷ reach answers "of the people who saw this, how many did something?" This is usually the one you want.
- Engagement ÷ impressions answers "per display, how often did something happen?" It runs lower, because repeat displays are in the denominator.
- Engagement ÷ followers answers something else entirely, and on any account where content travels beyond its followers it can exceed 100% without anything being wrong.
None of the three is incorrect. But an engagement rate quoted without its denominator is not a figure you can act on, and two tools quoting "engagement rate" are frequently not quoting the same thing.
A worked example, on a made-up account
Say a small bakery posts a photo of a new loaf. Over the week the post is displayed 4,000 times to 2,500 distinct accounts, and 300 people react, comment or share.
- Reach: 2,500. That is the audience size.
- Impressions: 4,000. So frequency is 1.6 — the average person who saw it saw it about one and a half times.
- Engagement rate on reach: 12%. On impressions: 7.5%.
Both engagement rates describe the same post honestly. If the bakery reports 12% one month and 7.5% the next because the tool changed its denominator, it will look like performance halved. Nothing happened.
The same arithmetic applies whether the account posts sourdough, personal-finance explainers, gym programming, gig listings or city-council coverage. The numbers are structural; the subject is irrelevant to them.
What to actually do about it
- Write down which metric answers which question for you, once, and stop switching. The discipline is worth more than the choice.
- Check the denominator on any engagement rate before you compare it to another one.
- Note the date a metric definition changed, and never compare across that line without saying so.
- When two sources disagree, do not average them. Find out which definition each is using; one of them is answering a different question.
None of this requires a particular tool. It requires knowing which of three similar-looking numbers you are looking at — which is, unhelpfully, the thing most dashboards are least explicit about.
In SupahPost the analytics surface keeps these separate and labels an estimate as an estimate, including the rate-per-thousand-views figure it calculates rather than receives. That is the same rule as everything above: a number is only useful if you know exactly what produced it.