Influencer Metrics: Reading Creator Numbers Without Being Fooled

What is engagement rate and how is it calculated?

Engagement rate is interactions divided by audience, as a percentage. The version this platform publishes:

Engagement rate v1 = (average likes + average comments) / followers × 100, averaged over recent posts.

That formula is deliberately limited to signals visible without account access, which keeps it comparable across creators — you can compute it for anyone, and so can we, and the calculator does. A creator's own analytics unlock a richer version including saves, shares and reach, which is one of the reasons claimed profiles here carry better data than unclaimed ones.

Two properties of ER that get forgotten:

• It shrinks as audiences grow. Bigger accounts nearly always show lower rates — distribution and audience dilution see to that. Comparing a 5k account's rate to a 500k account's rate tells you about tiers, not about the creators.
• It is an average with a distribution behind it. A creator whose rate comes from steady interaction across posts is a different proposition from one whose average is a single viral outlier — look at the spread, not just the number.

What is the difference between reach and impressions?

Reach is unique people; impressions are total displays. One person seeing a post four times is one reach, four impressions.

The distinction matters because the two inflate differently. Impressions grow whenever content is reshown — loops, revisits, feed re-serves — so impressions always read larger and always sound better in a report. When a report quotes only impressions, ask for reach; when frequency (impressions ÷ reach) climbs well above the low single digits, the same eyes are being counted repeatedly, which is fine for reinforcement and misleading for "we reached X people" claims. The glossary holds the compact definitions.

Which metrics actually predict sales?

In rough order of predictive usefulness for commerce:

1. Tracked conversions from past campaigns — the only direct evidence. A creator who can show disclosed results from a comparable brand is showing you the metric that matters.
2. Saves and shares — intent signals. Saving is deferred action; sharing is endorsement. Both cost the audience something, unlike a like.
3. Comment specificity — questions about price, sizing, availability are pre-purchase behaviour visible in public.
4. Engagement rate — a health check, not a sales forecast. It screens out dead audiences; it does not identify buying ones.
5. Follower count — a ceiling on reach and nearly nothing else.

That ordering is editorial judgement from how these signals behave, stated so you can disagree with it — not a measured ranking from a dataset we do not have.

What is a good engagement rate?

The honest answer: relative to tier and niche, or the question means nothing. Smaller audiences run structurally higher rates; niches differ in how interaction-prone their audiences are.

The orientation bands used in our calculator — around 1% weak, 1–3% typical, 3–6% strong, above 6% verify-the-inputs — are exactly that: orientation, published so they can be argued with, not a benchmark measured from a dataset. Use them to frame a question ("why is this account so far outside the band for its size?"), never as a pass mark. When this platform's index is large enough to publish observed distributions by tier and niche, those will replace the bands and be labelled as measured.

How do you spot fake followers?

You estimate — nobody outside the platform can prove it, and any tool claiming to definitively detect fake followers is overclaiming. The signals that converge:

• Engagement far below the plausible band for the size, or eerily uniform across posts.
• Comment quality: generic praise, emoji strings, the same rotating accounts within minutes of posting.
• Growth shape: vertical steps with no explaining event, or sawtooth rises and decays that track purchase-and-churn cycles.
• Audience geography that has nothing to do with the content's language or market.
• Ratios: following-to-follower patterns and like-to-comment balances far from the account's peers.

Each has innocent explanations alone; three together justify walking away. This is also why claimed profiles matter on this index — consented analytics narrow the guesswork, and our own audience-quality tooling will ship labelled as an estimate with its limitations stated, because that is what it will be.

Why does follower count mislead?

Because it is a stock, and everything a brand buys is a flow. The count records everyone who ever subscribed and never left — including the bored, the inactive, the bot residue of old growth tactics, and people whose interests moved on years ago. Reach per post, which is what a placement actually gets you, is routinely a modest fraction of the count and varies post to post.

The opinion this guide owes you plainly: follower count is the worst widely-used predictor of campaign performance and remains the first thing most brands filter by — including, statistically, whoever is reading this. Filter by it for tier and budget fit, since it does bound reach and price. Decide with engagement quality, intent signals and audience fit instead.

What should you measure for each objective?

Objective	Judge on	Ignore
Awareness	Reach, view-through, frequency	Same-day sales
Consideration	Saves, shares, profile visits, branded search movement	Raw impressions
Conversion	Tracked code/link sales over a window	Reach
Content production	Assets delivered vs licence	Engagement on her post

Two practical rules. Give each creator a unique code or UTM link, because shared attribution collapses the data. And judge conversion on a window of days or weeks, not launch day — purchase lag from creator content is real, and the search-then-buy path erases attribution unless you look for the branded-search lift too.

How often should metrics be refreshed?

Follower counts drift daily but decisions rarely hinge on the drift; engagement and reach patterns move on content cycles of weeks. Monthly refreshes are adequate for vetting, with a pull immediately before a contract is signed. What matters more than frequency is dating: every figure on this platform carries a collected-at timestamp, and every figure in your own reporting should too. An undated metric is trivia.

Metrics history matters as much as the snapshot — a creator at 4% falling from 7% and one at 4% rising from 2% are different bets at the same number. That is why profiles here store snapshots rather than overwriting, and render the trend once enough history exists.

What are story metrics versus feed metrics?

Ephemeral formats report differently: completion and exit rates replace likes, taps-forward and replies replace comments. The numbers run smaller and mean different things — a high completion on a fifteen-frame story sequence signals a strongly attached audience even if raw viewers look modest next to feed reach.

For paid work, the practical read: stories sell action (links, codes, direct response), feed and video sell durable association. Price and measure them accordingly rather than converting everything into one engagement number — which is also why our published ER formula does not pretend to cover story analytics it cannot see.

What to do this week

Take one creator you work with or follow, and build her picture from public data alone: ER via the calculator, comment quality on the last ten posts, growth shape, audience language. Write the one-paragraph verdict you would give a colleague. Then, if you are the creator: run the same exercise on yourself, date the numbers, and put them in the media kit the creator business guide tells you to build.