Should you compare creators by engagement rate?

Three names are left on the list. All three have been through the first check that asks nothing of anyone — you watched a few recent videos from each — and nothing on their public pages ruled any of them out. Now you have to decide who to write to first, and who is worth the slower checks.

Say the product is a desk fan quiet enough to run through a video call. The sentence the video has to land is you can leave it on during a meeting and nobody on the call hears it. All three of these creators film at a desk.

So you open a spreadsheet, give each name a row, and reach for engagement rate, because it sounds like the one reading that already accounts for how big the account is. Then you go to fill that column in, and there is nothing on the page to copy. It is a figure somebody has to work out, and there is more than one way to work it out.

Which engagement rate are you about to write down?

There is more than one formula, and the market has not settled on one.

One competitor's guide to calculating engagement rate on TikTok says so in its own words — "There are three different ways to find out your Engagement Rate" — and then gives three. The first divides the likes, comments and shares on a video by that video's view count. The second divides the same three by the creator's follower count. The third averages the likes, comments and shares across a run of recent posts and divides that average by follower count. The guide gives each one a job: the first for reading a single video, the second for loyalty, the third for an average over time and, in the same sentence, for comparing one creator with another, which is the question on this page.

Further down, past all three, the same guide sets out the method the vendor itself uses, which is that third shape: an average across a run of recent posts, over follower count. And the guide's own introduction states one formula up front without mentioning that the page will go on to name three.

Three isn't the whole range either. The same vendor's guide for Instagram, a different platform with its own screens, lists four, and two of those four divide by numbers nobody outside the account can open: impressions and a count of the separate people a post got to. Impressions, to be exact, count the times a post was put on a screen rather than the times somebody watched it.

So the word on your column header covers at least this much:

The last column is ours. Those guides hand over the formulas without sorting them by who can run them, and we have put that boundary down before. The guide on asking a creator for the numbers only they can see is about a different quantity, how far into an already-posted video people got, and it names this exact place: "There's something it was never sorting, and it's on a screen you don't have a login for."

The top two rows you can fill in for all three names this afternoon. The bottom two you can't fill in at all, because both of them sit in the account owner's analytics. They are in the table because the word on your column header stretches that far. Both were named for Instagram rather than TikTok, so neither is a formula for the platform your three candidates post on.

Of the two you can run, one is about a single video, and the guide on choosing creators has already said what that video's view count is: "The view count sits on each individual video, in public, one number per post. That one is about a single video and nothing else." One video is the unit everything below is about, so that is the formula this page's argument is built on. Whichever of the two you use, write down which one next to the number.

For the desk fan: you take the likes, comments and shares over the view count on the newest video from each of the three, and note the formula beside each figure. Two come out close to each other. One comes out at roughly twice the others, which is a fact about three videos and not yet a reason to put three people in an order.

Which two of these can go in the same column?

Only ones whose denominator you can see — in practice, ones you worked out yourself, or a set a tool computed one way and says so.

A percentage that arrives with none of that — a line in a media kit, a sentence in a reply, a column with no method printed next to it — doesn't bring its denominator with it. Two numbers divided by different things are not two readings of one quantity, and putting them in the same column doesn't make them one.

Two of those cases already have a guide between them. When the figures arrive as documents, what to do when a creator sends you a media kit rules on the pair of them, and it rests on the line our guide on asking a creator to send you their past work draws about figures a creator states in a reply: "The second one is a reason to keep those figures out of any comparison between two creators: each one came from a job you didn't see." What is different here is the object. A percentage isn't a document and isn't a claim about results. It is the output of a division, and the division is the part you can't see.

None of which makes a percentage you can't line up worthless. The choosing guide is careful about exactly that with the cumulative follower count, which it says is no kind of noise and is what you read every view count against, and the past-work guide is as careful about figures that arrive in a reply, which it says are no reason to discount what somebody sent you. A figure you can't line up is still a figure with a question attached, and the question is what it was divided by.

For the desk fan: one of the three already had an engagement rate against their name, from a tool your agency pays for, with no method printed beside it. The other two weren't in it at all. That figure goes in a column of its own with the tool's name at the top, your three go in another, and two cells in the tool's column stay empty because that is the accurate record of what it knows.

What does dividing do to the comments?

It turns them into a quantity.

The top of every one of these formulas is the same: likes, comments and shares, added up. Comments go in as how many arrived, not as what any of them said.

Our guide on telling whether a creator video worked is about a video you already paid for rather than a candidate you're still choosing, and it puts the value of a comment section somewhere a total can't reach: "The count is about people who saw the video. The comments are where some of them say what they took away, in their own words, right under it." It then splits them into three kinds — questions about using the product, descriptions of the product that get it wrong, and comments about the creator and nothing else — and reads that last kind, on a video made from your brief, as a verdict of its own: "It read as one of their posts, and your sentence never reached them."

Thirty comments about the creator and thirty questions about using the product put the same thirty on the top of the fraction.

For a candidate you haven't hired, the comments answer a narrower question, and reading them is the next check on the choosing guide's list: "Is the moment recognizable to the people watching? Read the comments under their newest videos." That same guide prices it at a few minutes with nothing asked of anyone, and says its own order isn't sacred, so the point isn't where the step sits. The point is that it answers something a count can't, and a ratio computed over the same comments doesn't answer it either.

For the desk fan: under the newest video from the creator whose figure came out twice the others, almost everything in the comments is about the desk — the lamp, the monitor arm, where the chair came from. Nobody asks how loud a fan is, because a fan wasn't what the video was about. The ratio counted all of it.

Does a high engagement rate mean this person fits?

No, and the reason was written down for a count rather than a ratio.

The guide on telling whether a creator video worked makes it about a video that has already gone up: "Take two videos sitting at the same number. One left people understanding that a sheet replaces the jug they're buying now. The other left them laughing at somebody falling into a washing machine. The number is identical and it settles nothing between them." The number there is a view count, and the reason given is that it was counting the audience rather than what happened to them. A ratio doesn't repair that. Underneath, it counts an audience. On top, it counts the reactions the section above showed it folds into a quantity. Neither end records what anybody took away, so two videos at the same percentage settle nothing between the videos. A figure worked out on one post was never a statement about the person who filmed it either.

What does settle fit has a name and a price. The first row of the choosing guide's table gives what a creator's recent videos can settle, "Whether the moment in your sentence has ever appeared in their work", and prices it at a minute with nothing asked of anyone. For the desk fan, that is whether anybody has ever filmed themselves on a call at that desk.

Which is why a cheap percentage still doesn't earn an early slot. The rule in that guide has two halves: "A check earns an early slot by being cheap and able to settle whether someone fits." The arithmetic is cheap. The second half is the one it fails, and the same guide has already written that failure out for the cheapest number on a profile, "the cheapest thing on the page to read, and it still can't settle whether the moment your sentence describes has ever appeared in someone's work." An engagement rate isn't that number. It is one you make by dividing what people left under a video by one of the two public numbers, and what the choosing guide's table puts opposite the question of fit is the videos themselves rather than either of them.

Should you ask for the numbers that would finish the formula?

Not to complete this column.

The guide on asking for the numbers only they can see answers yes to asking, and then says to ask for far less than you were about to: one named thing, in ordinary words, about a video they have already made for you. The thing on the other end there is a video you hired. Here it is somebody you haven't hired, and what you'd be asking for is a cell in a comparison table.

The choosing guide has already ruled on a check that needs the creator to send something, and its reason doesn't turn on what the check costs: "Even if this check were free it would still come last, because it can't tell you whether someone fits — only whether to trust what you're looking at." That is the same test the rest of this page keeps applying. Three figures off their own screens wouldn't settle fit either, whichever platform's formula you were trying to finish. Asking changes what you can put in the denominator, not what either end of the division counts.

So the asking buys three numbers whose limits you already know, from three accounts, and spends three asks on candidates whose free checks aren't finished.

For the desk fan: the message you were drafting asked all three of them for what their own screens say. You haven't read a comment section yet.

So what is an engagement rate for?

Pointing you at a video to open.

Take the first of the three formulas, the one that divides by a single video's view count, and run it over one account's recent posts a post at a time. Every figure comes out of the same division, on numbers anybody can read, so that run is the one comparison on this page nobody has to vouch for. The competitor's guide offers its own third formula for an average over time, so one account's record is already something the market asks this figure about. A post sitting well away from the rest of that run is the one to watch.

What the figure can't do once you get there was fixed somewhere else. The choosing guide read a profile's follower count against a view count and said what reading two numbers together produces: "They don't measure the same thing, so reading them side by side tells you where to look, not what to conclude." A run of one measure is the easier case and it arrives in the same place. The figure tells you where to look, and what you're looking for isn't in it: whether the moment in your sentence has ever shown up in this person's work, and what the people under that video were there for.

For the desk fan: you run the same formula over that creator's last several posts, and the newest one sits well away from the others. You open it, and it is a desk tour with a lot of people in the comments about the chair. That is a real finding about this creator, and the figure is what pointed at it. It isn't a reason to write to them first.

Frequently asked questions

What's a good engagement rate on TikTok?

There isn't a number here, and the reason is the one everything above turns on: a percentage can't be read until you know what it was divided by.

A threshold also has to be a threshold for something. A quoted average needs the set of accounts it was computed over, because that set is what decides what it is an average of. So check whether the page gives you both the set and the formula before you treat its figure as a line to clear. And fixing the denominator doesn't hand you a pass mark either. It makes the figure readable, which is a different thing.

Our tool reports an engagement rate for every creator on our list. Can we sort by that column?

Sorting it is fine. The order it produces is an order of percentages.

A tool that states its method and runs it over every name is the case this page allows, so those rows really are readings of the same quantity. What that buys is rows you can line up. It isn't an order of candidates, because lining figures up was never the step that settles fit.

And look at what the column is made of before you sort it. The method the competitor above publishes divides by the follower count, which the choosing guide describes as everyone who ever tapped follow and hasn't undone it, added up across a creator's whole history: "It doesn't describe last week." So the column orders your names by a recent average over a lifetime count.

See what this looks like for your brand

Start from the content you want, not follower counts.

Or email us: business@popow.ai