On 21 April 2025, Instagram made Views the primary performance metric across every content type — posts, Reels, Stories. It replaced impressions, and for Reels it absorbed plays and replays into the same figure.
The definition that matters: a view is counted any time the content plays on a screen, including when the same person plays it again.
Why this is not a cosmetic change
Before, a replay and a first play were different events, and some tools counted them separately. After, they are the same event. Which means the number got larger without anything improving, and it got larger unevenly — most for the content people watch twice.
Short, loop-friendly video benefits most. A long piece that nobody rewatches benefits not at all. So the change did not inflate everything by a constant; it redistributed apparent performance toward one format.
A metric that counts repeat viewing rewards looping. That is a product decision, and it is now inside your numbers.
The silent break in your own data
Here is the practical damage. If you compare a post from March 2025 with a post from May 2025, you are comparing a number that excluded replays with a number that includes them. The second will look better. It may not have been.
The same applies to anything built on a rolling history. A baseline computed over a window that straddles the change is a blend of two definitions, and every multiple measured against it is slightly wrong in a direction you cannot see. We wrote the repair separately: re-baselining after a platform changes a metric.
What this does to the case against view count
It strengthens it. Our argument has always been that view count mostly measures audience size rather than post quality — see seven metrics that beat view count. Views now also measures rewatch behaviour, which is a real signal, but it is mixed into the same figure with no way to separate it.
So the number is now a blend of three different things: how many people were shown it, how many watched, and how many watched again. A single number carrying three meanings answers no question cleanly.
Is the new metric worse?
Not exactly. For understanding distribution, views is more honest than impressions, because it counts actual plays rather than appearances on screen. And it aligns with what is reported to matter for ranking: watch time, which also accumulates across replays. In that sense views and the ranking signal now point in the same direction, which was not true before.
What it is worse at is comparison. It is less stable over time, less comparable between formats, and much less comparable across the April 2025 boundary.
What to do
Three things, in order:
- Do not compare anything across 21 April 2025 without saying out loud that you are doing it. If your window spans it, rebuild from after.
- Stop quoting raw views as a quality claim. It was weak before and it is noisier now.
- Watch the metrics the change did not touch — saves, sends, follows per reach. See saves and sends.
The general point
Platforms redefine their metrics, and they rarely restate history when they do. Anyone building measurement on top of a platform number is building on a definition that can move underneath them without notice.
That is an argument for writing your method down and dating it, which is why every method page here carries one.