A creator adds 70,000 followers in three weeks. The client sees the screenshot and wants them booked. Is that the best moment to buy this creator, or the worst?
You cannot answer from the follower chart. A vertical rise is the one shape that is genuinely ambiguous: it is what a hit post looks like, and what a purchase looks like, and what a cross-platform push looks like. The chart that resolves it is the one nobody screenshots: engagement volume over the same window, plotted underneath.
Read two curves, never one
Follower count is a cumulative number, and cumulative numbers hide their own derivative. It only goes up, it never expires, and it says nothing about whether the people behind it are still present.
Engagement volume is the corrective: total likes and comments per post, in absolute terms rather than as a rate. It is the closest public proxy for how many people are actually being reached, and unlike follower count it can fall.
Put both on the same time axis and the information is in the relationship between them.
- Followers up, engagement up in proportion: the audience is real and being served
- Followers up, engagement flat: the new followers are not present
- Followers flat, engagement rising: the existing audience is warming, or the format is being favoured
- Followers falling, engagement steady or rising: a purge or a cleanup, and often good news
The third and fourth of those are frequently the best creators on a shortlist, and they are invisible to anyone sorting by follower growth.
Five shapes and what they mean
Organic compounding. A gradual rise that steepens slightly over time, with visible steps where individual posts landed. After each step the line plateaus above the previous level rather than falling back to it. Engagement volume tracks the same shape with more noise. This is the boring, expensive, reliable creator.
A genuine viral spike with partial retention. A steep climb over days or a couple of weeks, then visible decay as incidental followers drift away, settling meaningfully above the pre-spike baseline. The decisive detail is that engagement volume rose during the spike and stayed partly elevated afterwards (below the peak, above where it started). The audience grew and some of it stayed.
A purchased spike. The same steep climb with no corresponding movement in engagement volume, followed by either a flat line or a slow bleed as the platform removes accounts. Engagement rate, being a fraction, collapses on schedule. The tell is the divergence, not the steepness; that comparison is the backbone of follower auditing.
Plateau and slow decline. Followers roughly flat or drifting down, with engagement volume falling faster than the follower line. This is a creator whose format has stopped working, or who has quietly reduced output. It is the most under-detected problem on shortlists, because the follower number still looks impressive and nothing about the shape is dramatic.
A platform purge. A sharp one-off drop, usually resolving within a day or two, with engagement volume undisturbed. Platforms sweep inauthentic accounts periodically, and a creator can lose several percent of their count without losing a single real reader. If engagement holds through the drop, they lost padding rather than audience. A creator who bought followers and was then swept shows the opposite: the drop follows an earlier rise that engagement never confirmed.
What each shape is worth
| Shape | Commercial reading |
|---|---|
| Organic compounding | Predictable delivery, priced fairly, rarely a bargain |
| Viral spike, retained | Elevated reach for a limited window; timing matters more than rate |
| Purchased spike | Avoid, and treat the media kit as unreliable on everything else |
| Plateau and decline | Negotiable, sometimes correctly, but falling engagement needs a reason |
| Platform purge | Usually neutral to positive; the remaining audience is cleaner |
The viral spike is mispriced in both directions
The interesting case is the creator four to eight weeks past a genuine hit, because both errors are common and they run opposite ways.
Overpriced. Rates get set against the peak. A creator who tripled their following in a month often raises their fee to match the new number while their per-post reach has already settled back toward, though above, where it was. You are buying a follower count that no longer describes the distribution.
Underpriced. The mirror case is a creator whose engagement rate looks poor immediately after a spike, because the denominator jumped and the numerator has not caught up. Any filter screening on engagement rate drops them. What that filter is detecting is recent growth, not a weak audience, and for a brand willing to read absolute numbers rather than the ratio this is the moment the creator is cheapest relative to what they deliver.
One question separates the two: has engagement volume, in absolute terms, settled above its pre-spike level? If it has, the audience genuinely grew and the rate will recover as the new followers are gradually re-served. If it has returned to the old baseline, the spike was traffic rather than audience, and the new followers are decoration.
There is also a timing consideration that has nothing to do with price. A creator riding elevated distribution will often deliver more in the following month than their historical median suggests, and less six months later once the platform has re-normalised them. Short campaigns can exploit that. Year-long ambassadorships priced off it will disappoint.
Why a single growth percentage hides all of this
"+18% in 90 days" is the number most tools display, and it collapses every distinction above into one figure.
It cannot tell you whether that 18% arrived in one afternoon or across ninety days, and those are different creators. It cannot tell you whether engagement moved with it. It is anchored to whatever date range the tool chose, so it silently rewards creators whose spike happens to fall inside the window. And because it is a ratio, a small account gaining 4,000 followers outranks a large one gaining 60,000.
Growth velocity is worth having. It belongs next to the shape, not instead of it. If you only ever see one number, ask what the chart under it looks like, and read it alongside engagement measured against the right peer group, because neither figure means much alone.
What the curves cannot tell you
- Public follower history is reconstructed, not recorded. Platforms do not publish a follower time series. Third-party charts, including ours, are built from periodic sampling, so resolution is limited and the date of a spike is approximate. Gaps get interpolated, and interpolation invents smoothness that was never there.
- A creator may have been tracked for less time than you assume. If an index first saw an account four months ago, its growth curve starts four months ago, and anything before that is invisible rather than absent.
- Engagement volume is also an estimate. Likes are hidden on some accounts, comments are deleted or filtered, and archived posts vanish from the record taking their numbers with them.
- Shapes overlap. A creator can go genuinely viral and buy followers in the same quarter. Real curves are messier than the five above, and two shapes at once is common.
- A clean curve is not a good creator. Everything here concerns whether an audience is real and present. Whether it is the right audience, in the right market, for this brand is a separate question, and usually the more expensive one to get wrong.
Where a chart helps and where you still have to decide
Software is good at the mechanical half: keeping a time series per creator, putting followers and engagement on one axis, and doing it across two hundred accounts rather than the five you would check by hand.
It is not good at the commercial call. Whether a creator four weeks past a spike is a bargain or a trap depends on your campaign window, your objective and what you are being quoted, and no chart resolves that for you.
Lyren tracks growth velocity per creator and shows it alongside engagement rather than as a standalone percentage, because the two apart are close to meaningless. The history is sampled rather than read from the platform, and it is presented as an estimate for that reason. Access is invite-only via the waitlist, starting with Romania.