What Is a Good Engagement Rate for Influencers?

Two people discussing “a 3% engagement rate” are almost never discussing the same thing. The numerator surprises nobody: interactions, meaning likes and comments. It’s the denominator that decides everything — and the denominator is the part nobody states.

Divide by followers and the rate answers one question: how much of my audience reacts? That’s the convention behind virtually every published figure, and it made sense when a chronological feed showed a creator’s posts to their followers, and roughly to them alone.

Divide by views and it answers a different question: of the people who saw this, how many reacted? That’s our choice, and it isn’t a matter of taste. On recommendation surfaces — the For You feed, Reels, Shorts — distribution no longer stops at followers. A video pushed to ten times its author’s audience shows a spectacular per-follower rate without anyone having done better work; the same video ignored by the algorithm collapses for the same reason, which has nothing to do with its quality.

• Per follower: “does my audience follow me?” — a measure of relationship.
• Per view: “did what was shown actually work?” — a measure of content.

Our percentiles are computed from our own collection runs, platform by platform and audience band by audience band, keeping only posts with a non-zero rate and accounts above a thousand followers.

We do not publish them here, for the same reason this whole article exists: a percentile taken out of its cell invites exactly the comparison you should avoid. They appear inside the product, next to the creator they describe — the only place where they mean anything.

We don’t publish an average. An average engagement rate is dragged upward by a handful of viral posts, at which point it describes nobody — least of all the median creator, who is the one you’re actually looking for.

• p25: a quarter of posts do worse. Below it, our product reads “poor”.
• Median: the middle post. From there up, “average”.
• p75: a quarter do better. Above it, “good”.

This is literally what the product does: the verdict shown on a profile compares the measured rate against its own cell — its platform, its audience band — and never against a universal constant. A good rate on YouTube would be mediocre on TikTok.

It’s the most repeated rule in the trade: small accounts supposedly engage better than large ones. Our records confirm it on Instagram and YouTube, where the median of the largest band sits at roughly two thirds of the smallest.

On TikTok, no. The median is close to flat from one end of the table to the other, and it actually rises again above five hundred thousand followers. Which is consistent with the denominator: if distribution is decided by recommendation rather than by follower count, audience size stops being the dominant factor.

Another received idea our sample does not support: the ordering “TikTok, then Instagram, then YouTube”. TikTok leads comfortably at every band. But Instagram and YouTube are neck and neck between ten and fifty thousand followers — separating them on this measure alone would be reading noise.

And here is where the choice matters most: Instagram’s ten-to-fifty-thousand band carries the widest upper tail in the entire table, its p75 sitting at more than three times its median. Two creators of identical size can be a factor of three apart. That is precisely the band where looking at the rate instead of the follower count changes the outcome of a campaign.

A real case, taken from our own code: a post with ten views and three interactions shows a 30% engagement rate. Statistically, that number says nothing at all.

So we enforce a sample floor. Below a hundred views the rate is not displayed: it is treated as unknown, the interface shows a dash, and the relevance engine scores it zero. Before that floor existed, dormant and test accounts surfaced at the top of the engagement sort — the ranking was rewarding noise.

It’s also a debatable decision, and worth saying so: it hides information about accounts that are real but have very little reach. We preferred withholding a number to presenting it as comparable. An unknown rate is never displayed as a zero.

• They do not compare to per-follower benchmarks, however similar they look. A 3% here and a 3% elsewhere are not measuring the same thing.
• They come from three verticals — beauty, food, tech. A fourth could move the cells; fashion and gaming in particular are absent.
• They describe posts, not campaigns. No link to a click-through rate, a signup or a sale is established here, and we claim none.
• Two cells rest on few observations, both on Instagram: between fifty and a hundred thousand followers, and above five hundred thousand.
• They will be re-measured as our volume grows. The calibration date is displayed so you know what you’re reading, and from when you should start doubting it.

BriefBench places every creator against their own platform and audience band, with the measured rate and the verdict that follows from it — not a universal constant.