The list comes back from the client with three names crossed out and a fourth added by their marketing manager, who saw someone on TikTok last week. The creators you put forward were not worse than the one that replaced them. They simply arrived as names, and a name has nothing to defend itself with.
A shortlist gets argued with when it reads as a ranking somebody produced, and gets approved when it reads as a set of decisions somebody made. The difference is almost never the quality of the creators.
A shortlist is a portfolio, not a ranking
Ranking implies a single axis, and there isn't one. Ten names sorted by anything is an open invitation to ask why number four sits above number five: a question you cannot win, because the honest answer is that they are not really comparable on the axis the sort implies.
Compose the list in groups instead, and label the groups.
- Anchors. One or two larger creators, carrying reach and carrying credibility with the client's own internal stakeholders. They are usually the least efficient names on the list, and you should be the one who says so.
- The engaging middle. The bulk of the list, and where the campaign's actual performance tends to come from. This is where the follower-tier decision gets made deliberately rather than by drift, because shortlists left alone drift upward: big numbers feel safer in a deck.
- One or two bets. A creator who is a stretch on niche, size or format, included for a specific stated reason. Label them as bets. A named bet that underperforms is a decision that did not pay off; an unnamed one is a mistake.
A client argues with a list of ten names. They engage with three anchors, five in the middle and two bets, because that structure invites a conversation about strategy rather than about individuals.
The error nobody checks for
Audience overlap. Three creators sharing the same audience are one creator with three invoices, and this is the most commonly missed failure in an otherwise careful shortlist.
It gets missed because every name checks out individually. Overlap is a property of the set, not of any member of it, so a process that evaluates creators one at a time is structurally unable to see it.
You can get most of the way there without tooling.
- Look for mutual presence. Do they appear in each other's comments, tag each other, appear in each other's content? Adjacent creators in a small niche usually do, and that is your first flag.
- Check for clustering. Same sub-niche, same city, same format, similar size. Four fitness creators in the same city making the same kind of content are usually selling access to the same room.
- Sample the follower lists. Open two of them and look for the same accounts appearing in both. Crude, unrepresentative, and still worth the ten minutes it takes.
- Ask about the client's own history. If two of your picks both ran for this brand last year, somebody has already measured the overlap and the results are sitting in the client's reporting.
Some overlap is fine and occasionally it is the point, when the plan is built on frequency rather than reach. Unplanned overlap is the version that quietly halves a budget's effect while every individual line item still looks defensible.
Write the reason and the risk
For each creator, one or two sentences: why they are on the list, what evidence supports it, and what the risk is. It looks like this.
Skincare, mid-tier, Instagram. On the list because their routine posts consistently out-perform their own median, and the audience reads as domestic on every check we ran. Risk: they ran a competitor's campaign fourteen months ago, and posting has been irregular since June.
Two things change when the list looks like this. The client stops evaluating names and starts evaluating reasoning, which is what they are actually paying for. And you keep a record you can review in six months, when the campaign is being assessed and nobody remembers why the fourth name was there. Follower-authenticity work belongs in that evidence sentence rather than in an appendix: the signals worth citing are here.
Stating the risk yourself is the part that buys credibility, and it is the part people cut when they are short of time. A list with no risks on it does not read as a list with no risks. It reads as a list nobody checked.
The format is part of the argument
A link to a spreadsheet is not a deliverable, and the reasons are practical rather than aesthetic.
It does not survive being forwarded. The client's marketing manager sends it to a brand director who opens it on a phone, sees a grid of columns, and forms an opinion about the agency before reading a single reason. Spreadsheets also invite sorting, and the moment somebody sorts by follower count the portfolio you composed collapses back into the ranking you deliberately avoided. Then access breaks: permissions, versions, the copy somebody downloaded in March and is still quoting from.
Give them a document instead. A PDF or a shareable page that keeps the reasoning next to the name, holds the order you chose, reads on a phone, and forwards without a permissions conversation. Most of the perceived value of the research lives in this last step, which is exactly why leaving it to the final twenty minutes is such an expensive habit.
The client who wants the biggest name
This request is common, and from their side it is reasonable. Do not refuse it. Price it.
Put the large name on the list as an anchor, with its cost stated as a share of the total budget and with what the same money would buy in the middle of the list set beside it. Being the person who included it and named the trade-off is a much stronger position than being the person who left it off and has to justify the omission.
Then ask what the name is actually for. Frequently the objective is internal: a board deck, a stakeholder who needs to recognise somebody, a competitor who ran a similar campaign. That is a legitimate objective, and knowing it changes which large name is the right one: it may not be the biggest.
If they still want it against the evidence, write the recommendation down and run the campaign well. A documented disagreement is worth considerably more at the post-mortem than being quietly right was.
The shortlist is also the input to the next document. Knowing which of a creator's themes already outperform their own baseline is what separates a brief that helps from one that flattens the creator, so what belongs in the brief is worth settling before the approval lands rather than after.
What a good shortlist still cannot promise
It makes a decision reviewable. It does not make it correct, and the difference is worth stating to the client rather than letting them discover it.
- It does not predict performance. Every signal available from outside is historical and public. A creator with a clean record and a well-matched audience can still make a post that lands flat.
- Overlap checks from outside are approximate. Without both creators' own analytics you are inferring shared audience from public surfaces, and the true figure could be well above or below what you found.
- Reasons can be confident and wrong. Writing them down makes them reviewable, not accurate. That is still the point: a wrong reason you can find in six months is more useful than a right instinct you cannot reconstruct.
- The risk you state is only the risk you found. Brand-safety scanning from public data misses private accounts, deleted posts and anything said off-platform.
- A well-argued list can still lose to a preference. Documenting the reasoning is the goal, because it protects the relationship when the preference does not work out. Winning the meeting is not the same thing.
Lyren produces the input to this step rather than the step itself: a ranked candidate pool with the peer comparison already done, plus content and posting analysis to draw the evidence sentences from, exported as a PDF or shareable report (white-label on agency plans) so the deliverable is not a spreadsheet link. Composing the mix and writing the reasons stay with you, because that is the part the client is paying a person for. Access is invite-only from the waitlist.