Influencer discovery gets cheaper when you qualify in the right order

Finding candidates is easy and takes an afternoon. Deciding which of them to actually approach is the part that takes judgment, and it is the part most teams rush.

Discovery ends when you have a small number of names you would defend in a meeting, with a written reason next to each. Everything before that is collecting.

Cost order is what makes that cheaper. The first mechanical check takes about 30 seconds per candidate. The same candidate read at the comment stage takes about four minutes. That gap is the argument for running the list in cost order.

What to take away

The objective narrows the search before you start

If the campaign's purpose is unsettled, discovery returns people who are simply popular. That is the default output of every search tool ever built.

The objective and the constraints around it make the first cut. They decide whether you seek audiences you do not have, credibility in a specific subject, or people whose followers already buy this way.

Those three searches return different names. Running one search and hoping it serves all three is how a shortlist ends up full of accounts nobody can quite explain.

Where candidates actually come from

The last row deserves names: SparkToro for audience overlap research, Social Blade for public account history, and Modash, HypeAuditor or CreatorIQ for filtering a list at scale. Each covers one or two routes above, not all seven.

Typical self-serve prices run from a free tier to a few hundred dollars a month. Enterprise creator platforms are usually quoted annually and per seat, so ask for the per-seat figure before you build a process around one.

Use at least two routes. Any single one has a shape, and a shortlist assembled from one route inherits that shape without anybody noticing.

The first row is the most underused. Customers who already post about the category are the cheapest strong candidates available to most brands, and they arrive with something no search can produce: genuine prior use.

Qualify in order of cost

Run the cheap checks first and drop candidates as early as possible. Nothing wastes a week faster than doing careful work on someone who was never eligible.

Only the last two need a person thinking. The first five are mechanical and disqualify most of a long list.

Work down this list and stop at the first failure. Times are typical per candidate, with the brief open in front of you.

Mechanical checks

Judgment checks

The arithmetic is the point. Ten candidates that fail check 1 cost about five minutes in total. The same ten taken as far as the comment read cost about two hours.

Each of those ten had already been through the other five checks. A candidate who clears all seven has cost you roughly 20 to 25 minutes, so the order matters more than the tooling.

Reading the audience without fooling yourself

Follower counts are cheap to compare, which is precisely why they dominate decisions they should not.

Treat any breakdown of who an audience is as an estimate rather than a measurement, unless it came from the creator's own platform analytics for a stated date range. Third-party breakdowns are built from a sample of visible followers, so they carry a margin nobody displays; the underlying idea is sampling error.

Estimates are useful for excluding candidates and sorting a list. They are not useful for close calls: two candidates a few percentage points apart on an estimated figure are not separated by anything.

Where a specific audience characteristic really matters to the deal, ask the creator for a screenshot of their own analytics covering a period you name, and keep it with the file. That is a measurement, and it puts the creator on record about it.

Growth shape is worth a look. An account that gained its audience in steps, with flat stretches between, is telling you something different from one that grew on a curve. Neither is proof of anything on its own, and both are worth a question rather than an accusation.

There is a real market in manufactured audience indicators, and it has drawn enforcement as well as industry complaint. The FTC has brought cases over the practice. One is the Devumi case, set out in a 2019 press release about the sale of fake indicators of social media influence, which names the company, its owner and its chief executive.

Comments are the cheapest signal available

Open thirty recent comments on three recent posts and read them properly. It takes about four minutes per candidate and it tells you things no metric will.

An audience that argues with a creator is usually a better audience than one that only agrees, because argument means people are actually reading. A comment section full of single emoji and nothing else is a comment section that will not produce a purchase, whatever the engagement rate says.

Building a shortlist you can defend

Aim for a number you can genuinely write about, one paragraph each: who they are, why this campaign specifically, what you are unsure about, and what you would need to see to be sure.

That last field is the one worth insisting on. A shortlist where every entry looks perfect is a shortlist that has not been examined. Recording the doubt makes the eventual conversation better, because you know what to ask.

Keep the rejections too, with the reason and the date. Half of them are wrong within a year: exclusivity expires, an account changes direction, a small creator grows into the brief. A rejection list with reasons is a search you never have to run again.

Where discovery stops

Discovery hands over a shortlist. It does not verify identity, agree deliverables, negotiate rights, or establish that anyone is available. Those are separate jobs with separate failure modes, and treating a shortlist as a plan is how campaigns end up with four names, none of whom ever agreed to anything.

The handover is worth making explicit. Whoever does the approaching should receive the reasoning, not just the names, or the first message will be generic and the work you did will be invisible to the person you did it for.

Bottom line

Let the objective cut the field before you search, use more than one sourcing route, run the cheap disqualifying checks before the expensive judgment ones, and read the comments. Then write a paragraph per name, including what you are unsure about.

Common questions

How many candidates should a shortlist have?

Enough that a no does not stall the campaign, few enough that you can write a real paragraph about each. If the paragraphs start repeating, the list has stopped being a shortlist.

Do smaller accounts perform better?

Sometimes, and the reason is usually specificity rather than size: a smaller account often serves a narrower audience and answers its comments. Size is a proxy for those things, not a cause, so check for the things themselves.

Should we rule out anyone who has promoted a competitor?

Not automatically. Check when, how recently, and whether any exclusivity is still running. A creator who works across a category can be a stronger endorsement than one who has never mentioned it, though your own view on that should be decided in advance rather than in the moment.

What if the best candidate has no commercial route?

Treat the absence as information rather than an obstacle to route around. They may not want commercial work, or may be represented in a way you have not found. Both answers are worth having before you spend time.