Ask an agency strategist how they build an influencer shortlist and you will get an embarrassed version of the truth: platform search, a lot of tabs, a spreadsheet, and judgement applied at the end when everyone is tired. It takes a day or two per brief. It is repeated from scratch every time. And the quality of the output depends almost entirely on how fresh the person was when they got to the last twenty profiles.
This is not incompetence. It is what the available tools make possible. But the process has a structure, and once you see the structure you can do it faster and better.
Where the two days actually go
Very little of it is thinking.
- Searching. Platform search is built for finding a person you can already name, not for finding people who match a description. So you search a hashtag, scroll, open promising accounts, and repeat across three platforms with three different interfaces.
- Copying. Handle, follower count, a guess at engagement, a link, into a spreadsheet. Per creator. Times eighty.
- Checking. Is this person actually in the right market. Are those followers real. Have they posted this month. Did they work with a competitor last quarter.
- Formatting. Turning the spreadsheet into something a client will read.
The thinking — which creators fit this brand, which are worth their rate, what the mix should be — happens in the gaps, and gets the least time.
The five decisions that determine whether the shortlist is good
Everything else is execution. These are the calls that matter, and they should be made before the searching starts, not discovered halfway through.
1. Audience market, not creator market
The single most expensive error in smaller European markets. A creator based in the country is not the same as a creator whose audience is in the country, and in a market with a large diaspora those two numbers diverge badly. Decide which one the brief actually needs and write it down. The full case is here.
2. The follower band, chosen from the objective
Reach objectives and conversion objectives want different bands, and the honest version is that smaller accounts usually engage harder while larger ones deliver volume more predictably. Pick the band from what the campaign is for, then hold it. Shortlists drift upward because big numbers feel safer in a client deck.
3. Niche precision versus adjacency
"Fitness" is not a niche; it is four or five. Strength training, home workouts, running, yoga and physique content share a label and share almost no audience behaviour. Decide whether you want the core of a narrow niche or the edges of a broad one, because that decision changes the entire list.
4. What "good engagement" means for this brief
An engagement rate is meaningless without its peer group. Comparing a creator to a global average tells you nothing; comparing them to the median for their niche, platform, market and size tells you a great deal. Fix the comparison before you start, so you are not re-litigating it per creator.
5. Brand safety scope
Decide up front how far back you read and what disqualifies. Doing this per creator, ad hoc, at 6pm, produces inconsistent decisions you cannot defend later.
The process
Step 1 — Convert the brief into filters
Turn the client's language into something checkable. "Aspirational but relatable, 25-40, urban" becomes a follower band, a niche list, an audience-market threshold, and an activity requirement. If a criterion cannot be turned into something you could verify on a profile, it is a tiebreaker, not a filter. Say so.
Step 2 — Build a long list wide, and fast
Cast wider than feels necessary — three to five times the final shortlist. The purpose is coverage, not quality; quality comes next. Two failure modes to avoid: searching only the platform you personally use most, and stopping at the first twenty acceptable names, which biases the whole list toward whatever the search surfaced first.
Step 3 — Cut on the hard filters, without opening anything
Market, follower band, last-posted date, niche. This is mechanical and should be ruthless. Do not open profiles yet; opening profiles is what turns a filter into a two-day job.
Step 4 — Score what survives
Now you look properly, and only at a list you have already cut to a manageable size. For each: engagement against the peer median, what actually performs in their content, posting consistency, growth direction, and the audience-authenticity signals. The signals that survive a careful seller are here.
The output of this step is not a ranking. It is a set of creators with a reason attached to each.
Step 5 — Compose the mix
A shortlist is a portfolio, not a top ten. A good one usually has a small number of anchor creators for reach, a larger group in the band that engages hardest, and one or two deliberate bets. Overlap matters here: three creators with the same audience are one creator with a bigger invoice.
Step 6 — Write the reasons down
For each creator, one or two sentences: why they are on the list, what evidence supports it, what the risk is. This is the difference between a list a client accepts and a list a client argues with. It is also what makes the shortlist reviewable in six months when the campaign is being evaluated.
Step 7 — Deliver something durable
A link to a spreadsheet is not a deliverable. A PDF or a shareable page with the reasoning attached survives the meeting, gets forwarded internally, and can be revisited. Most of the perceived value of the research lives in this step, which is why it is such a mistake to leave it to the last twenty minutes.
Where tooling helps, and where it does not
Software is genuinely good at the parts that are mechanical and comparative: building a candidate pool across platforms without manual searching, applying hard filters, and — the one that is impossible by hand — comparing a creator against the right peer group, because that requires a distribution you cannot assemble yourself.
It is not good at steps 1, 5, and 6. Converting a brief into filters is interpretation. Composing a mix is judgement about a specific brand. Writing the reasons is the work you are actually being paid for.
Be suspicious of any tool that claims to do those three. The useful version compresses steps 2, 3 and 4 from a day and a half into minutes, and hands you back the time to do the parts that need a person.
A note on database size
The obvious question when comparing tools is how many creators are indexed. It is close to the wrong question, and worth its own discussion. A hundred million creators you cannot filter accurately by audience market is not a hundred million candidates for a Romanian brief. Coverage of your market and accuracy of the filters beat raw index size for every brief that is not global.
Lyren is built around this workflow: niche plus audience market plus platforms produces a ranked candidate pool with the peer comparison already done, so the time goes into steps 1, 5 and 6 instead of step 2. It is invite-only while we build out market coverage, starting with Romania.