Two creators land on the same shortlist. Both Instagram, both around 40,000 followers, both in the client's market. One reports 1.4% engagement, the other 3.9%. The shortlist keeps the second and drops the first.
That was closer to a coin flip than to analysis. Nothing in those two numbers tells you which creator is performing well for what they do, because engagement rate varies more across niche, follower tier and platform than it does between a good creator and a bad one. A rate without its peer group is not a measurement. It is a digit.
The rate is not one formula
Ask three tools for a creator's engagement rate and you will often get three answers, none of them wrong. They are computing different things.
The numerator is usually likes plus comments. Some tools add saves and shares where the platform exposes them, which can lift a rate noticeably for content people bookmark — recipes, workouts, checklists — and barely move it for content people scroll past approvingly.
The denominator is where the real divergence lives.
| Denominator | What the rate then means | Where it breaks |
|---|---|---|
| Followers | Share of the subscribed audience that reacted | Useless when most distribution goes to non-followers |
| Reach | Share of the people who saw it that reacted | Only the creator sees reach; third parties estimate it |
| Views or plays | Share of plays that produced a reaction | Not comparable to follower-based rates, and a view is a low bar |
A media kit that says "4.2% engagement" without saying against what is not disclosing a metric. It is quoting one. Before comparing two creators, confirm both numbers came out of the same formula, or you are comparing a fraction to a different fraction.
Follower tier moves it more than quality does
Engagement rate tends to fall as accounts grow, reliably enough that comparing a 15,000-follower creator to a 900,000-follower creator on this metric tells you almost nothing about either.
Several things drive it at once. A small audience is usually self-selected and genuinely interested; a large one accumulates people who followed for a single post two years ago and never came back. Feed ranking gives each post a limited early sample regardless of audience size, so the ceiling on reactions does not scale linearly with followers. And follower counts only ever go up, while attention does not — dormant accounts sit in the denominator forever.
The practical consequence is that a follower band is not a nice-to-have on a brief; it is part of the measurement. "Engagement above 3%" applied across a list spanning 10,000 to 500,000 followers does not select for quality. It selects for small.
Why the niche median is the only comparison that decides anything
Niches differ in how much reacting they ask of an audience. Humour, fandom and meme content collect likes almost as a reflex. Finance, B2B, legal and medical content is read carefully by people who do not tap a heart on the way past. Neither audience is better. They behave differently, and a single benchmark applied to both will systematically flag one group as excellent and the other as failing.
That is the category error underneath most published engagement benchmarks: averaging across niches produces a number that describes no niche.
The comparison worth making holds four things constant.
- Platform — an Instagram rate and a TikTok rate are different units, not different values
- Niche — narrow enough to be real; "fitness" is four or five niches under one label
- Follower tier — a band, not an order of magnitude
- Audience market — reaction norms are cultural as well as topical
Against that peer group, a percentile is worth more than a percentage. "Above the 80th percentile for beauty creators in Romania at 20,000 to 60,000 followers" is a finding. "3.1%" is trivia.
Short-form broke the denominator
Follower-based engagement rates assume distribution follows subscription: you post, your followers see it, some react. Reels, Shorts and TikTok's For You feed do not work that way. They distribute to people who do not follow the account, and on TikTok that is the default rather than the exception.
Two things follow. A follower-based rate on short-form content can exceed 100% without anything being wrong — it means the post travelled. More damaging, the metric stops describing the creator's audience at all and starts describing one post's algorithmic luck.
For short-form the useful denominator is views, and the useful reading is a distribution rather than a point. Look at the median view count over the last twenty to thirty posts, look at the spread, and look at views against follower count. A creator whose views run consistently well above their follower count is being pushed. One whose views run consistently below it has an audience the platform has stopped serving — one of the sharper fraud signals, though it has innocent explanations too.
What the published averages are actually for
They are a sanity check, and nothing else.
Published industry studies for 2026 put typical Instagram engagement somewhere near 1% for static posts and higher for Reels, TikTok in the region of 4% when measured against views, and YouTube in the low single digits. Those are aggregates from third-party panels, computed with formulas the studies do not always state, and they mix every niche, every follower tier and every market into one figure.
Use them the way you would use a rule of thumb about average salary: to notice that something is off by an order of magnitude, never to price an individual. If a creator's reported rate is 0.05% or 25%, a published average tells you to go and look. Between those poles it tells you nothing.
Treat the year on the study as load-bearing, too. Distribution changes, formats change, and a 2023 benchmark applied to 2026 short-form is measuring a different internet.
Setting the comparison before you search
Deciding what "good" means per creator, at the end of a long day, produces a shortlist you cannot defend. Decide it at brief stage, as part of the wider research workflow.
- Fix the formula. Choose one numerator and one denominator per platform and apply it to everyone. Reject media-kit numbers computed some other way, or recompute them yourself.
- Fix the peer group. Platform, niche, follower band, audience market. Write it into the brief so the threshold travels with the shortlist.
- Express the threshold as a position, not a value. "At or above the niche median" survives a change of year, format and platform. "Above 3%" does not.
- Read consistency alongside level. A rate that never varies is a stronger warning than a rate that is merely low, because real audiences produce hits and misses.
- Put the rate next to the growth curve. Engagement per follower falling while followers climb is among the most informative patterns in creator vetting, and it is readable from two charts.
One consequence worth stating: because engagement only means something against a peer group, any composite creator score that folds it in has to be scored within a niche too, or it inherits the same category error. What such a score should actually measure is a fair question to ask of any tool that shows you one.
What an engagement rate cannot tell you
Even computed correctly and compared correctly, this metric answers a narrow question.
- It does not measure intent. Likes are not purchases. An audience that reacts warmly to everything and buys nothing clears every threshold here.
- It does not measure audience fit. A genuine, engaged audience in the wrong country is a more expensive failure than a mediocre rate, and a different check entirely.
- It rewards engagement bait. Comment-to-win, tag-three-friends and rage-bait formats lift the number without lifting anything a brand wants. Read the comments before you trust the rate.
- It is an outside estimate. Nobody outside the platform sees reach or impressions. Every third-party rate, including ours, is inference from public counters, and inference on small accounts is thinner than the decimal places suggest.
- Medians are estimates too. A niche median is only as good as the sample of creators behind it. In a small market and a narrow niche that sample can be a few dozen accounts. Treat it as a guide, not a constant.
Lyren scores engagement against the niche median, within a follower band and an audience market, rather than against a sitewide average, and shows where a creator sits relative to that peer group instead of handing you a bare percentage. Where the peer sample is thin, that is worth saying out loud, and we would rather say it than round it away. Access is invite-only via the waitlist while market coverage is built out, starting with Romania.