A Romanian travel creator with 62,000 followers on Instagram reports a 0.7% engagement rate. A marketer running down a shortlist reads the number, assumes the audience is inflated or asleep, and crosses the name off without opening the account.
Nothing about that number says fraud on its own. It could be the format, it could be timing, it could be a real audience that is simply the wrong fit for this particular brief: three different problems that produce the same low digit and call for three different responses. What counts as a good engagement rate is a peer-group question with a defensible answer. What a number below that peer group actually means is a separate question, and treating every low reading as the same problem is how good creators get cut and genuinely weak ones survive the same shortlist.
Five reasons a number reads low and nothing is wrong
- The wrong denominator for the format. A follower-based rate assumes people see a post because they follow the account. Long-form YouTube and static Instagram posts still mostly work that way; TikTok and Reels largely do not, because the majority of views come from people who have never followed the creator. A short-form account judged against followers rather than views will read low by construction, regardless of how the audience actually behaves.
- Account size. Engagement rate falls, reliably, as an account grows. Large audiences accumulate followers who reacted once, years ago, and never opened the app again: they never leave the denominator, and feed distribution does not scale linearly with follower count to compensate. A 400,000-follower account reading lower than a 15,000-follower account on the same content is closer to the default outcome than the exception.
- Posting cadence and window. An account that posts daily and one that posts a few times a month are not comparable on a raw rate calculated the same way over the same window. A busier feed spreads a similar total amount of attention across more posts, thinning the rate per post without the creator's actual reach having changed.
- A quiet stretch, not a quiet audience. Distribution algorithms suppress and boost unevenly, and a two- or three-week dip in reach can pull down a thirty-day average even when the account's typical weeks look nothing like it. A single low window is weak evidence on its own: the shape across several months tells you more than any one snapshot.
- Audience-market mismatch, not audience quality. Audience location and creator location are different numbers, and a genuinely responsive audience sitting in the wrong country can look, on the raw rate, identical to a small or disengaged one. The engagement is real. The brief was written against the wrong variable.
Three signals the number really is a warning
Not every low reading is explainable. A smaller set of patterns is worth treating with real suspicion rather than benefit of the doubt.
- Sustained, not situational. A rate that stays low across months and across formats, rather than dipping once and recovering, points to something structural about the audience rather than an unlucky week or an algorithm hiccup.
- Reactions without conversation. Likes accumulating with almost no comments, or comments that read as generic and interchangeable across unrelated posts, is a pattern worth reading closely at the account level rather than taking the aggregate rate at face value.
- Inconsistency the audience should not produce. A creator whose TikTok performs normally and whose Instagram reads dead, on similar content and a similar audience size, has a platform-specific problem (bought followers on one account, a distribution issue on the other, or genuinely different audiences on each) that a single blended rate hides rather than reveals.
When the number is accurate and the fit is still wrong
There is a third bucket that is neither fraud nor a measurement artefact: a real, healthy account whose content simply does not land with the audience a particular brief needs. A finance creator with strong engagement among people who already invest is not underperforming. They are the wrong creator for a campaign aimed at people who have never opened a brokerage account, and nothing about their engagement rate is misleading: it is honestly describing an audience that is not the one being bought for this particular product.
This is the case a pure threshold cannot catch, because the number itself is genuinely good. It only becomes visible once you look at what the engagement is actually about (the comments, the recurring content themes, who is replying) rather than how much of it there is. A rate this healthy passing every numerical filter and still producing a campaign that underperforms is usually this problem, not a fraud problem.
Reading the number before you act on it
- Confirm the denominator matches the platform. A follower-based rate on short-form content is close to meaningless on its own; check it against views before drawing any conclusion.
- Look at a spread of posts, not one average. Recent months at a glance separate a structural pattern from a bad stretch that has already passed.
- Read fifty comments before reading the percentage. The rate tells you how much reaction there was. The comments tell you what kind, and whether it looks like a real audience talking.
- Compare against the peer group before comparing against a rule. A number that looks low against a flat threshold can sit comfortably inside its own niche and follower band once that band is the actual comparison, and an authority score built within the niche is doing the same comparison automatically rather than one number at a time.
What none of this can tell you
- It cannot separate a quiet audience from a small one. Both produce a low rate, and only the growth curve and the comments reliably tell them apart.
- It cannot tell you whether reactions convert. A rate above every benchmark in this article still says nothing about purchase intent, which is a different measurement entirely.
- It is an estimate on both sides of the comparison. The creator's rate and the niche median it gets compared against are both inferred from public counters, not measured directly, and a thin sample makes the comparison shakier than the decimal point suggests.
- It does not replace reading the content. A single number is a summary of a few hundred data points that a person could, and sometimes should, look at directly.
Lyren reads engagement against the niche median, within the same platform, follower band and audience market, rather than against one published threshold, and surfaces a creator's actual posting pattern alongside the score instead of leaving a low number unexplained. Where the peer sample behind a median is thin, that gets flagged rather than smoothed into false confidence. Access is invite-only via the waitlist, starting with Romania.