Engagement rate is the share of people who did something with your post — liked, commented, shared, saved — out of the people who could have.
The formula looks simple:
engagement rate = (likes + comments + shares) / views
The trouble starts with the divisor.
Nobody agrees on the divisor
Some tools divide by views. Some divide by followers. Some divide by reach. The same post can score 2% or 12% depending on which one a dashboard picked, and most dashboards don't tell you.
| Divisor | What it answers |
|---|---|
| Views | Of the people who saw it, how many acted? |
| Followers | Is my audience actually paying attention? |
| Reach | How did it do with unique humans? |
For a founder, views is usually the honest one. Followers-based rates flatter small accounts — 10 likes from 100 followers is 10%, and it means very little.
Half the inputs don't exist
This is the part tools rarely admit:
- Instagram never exposes saves to anyone but the account owner. Any "engagement rate" that claims to include saves on someone else's post is guessing.
- Photos have no view count on Instagram — so a photo's engagement rate can't use the same formula as a Reel's.
- Shares are owner-only on most platforms, including YouTube and Reddit.
A missing input should produce null, not a zero. A tool that silently treats "no data" as "zero engagement" will tell you a post flopped when it simply couldn't measure it.
Why a high rate can still mean nothing
Engagement rate is a ratio, and ratios hide size. A post seen by 40 people with 8 likes scores 20% — better than anything you'll ever get at scale. It also reached 40 people.
Read the rate next to the raw reach, always. On its own, it rewards being small.
What to do with it
Use engagement rate to compare your own posts to each other, on the same platform, over the same period. That's the comparison where the divisor is consistent and the number means something.
Comparing your rate to someone else's, or to an "industry average", is comparing two different formulas run on two different sets of missing data.