Discovery platforms strategy explained for 2027

Buying a search product is not a strategy. The strategy is what happens between a brief arriving and a shortlist going out: how the brief becomes a set of conditions, where the first candidates come from, how the list widens, and what tells you to stop. Teams that have written this down find people faster with a modest tool than teams that have not find with an expensive one.

What to take away

• Search for the audience first and the creator second. Most briefs describe a customer and most searches describe a person, and the gap between them is where the shortlist goes wrong.
• A seed set beats a filter set. Start from a handful of people you know are right, then work outwards from what they have in common.
• Your exclusion list is the part that compounds. It is worth more after two years than any index subscription.

Turn the brief into conditions before you open anything

A brief says who the product is for, what it does, and what the campaign has to achieve. A search box wants categories, follower ranges and locations. The translation is a real piece of work and it should be written down, because it is the thing you will argue about later.

Three questions do most of the translation. Who is the content actually meant to reach, described as people rather than as a segment name. What would a creator have to be doing already for this brief to make sense on their channel. What would make a technically qualified creator wrong for it anyway.

That third question produces the conditions no filter offers: tone, the kind of comments the account attracts, whether the person has taken money from something you cannot sit beside. Write them as sentences. They will be applied by a human at the shortlist stage, and if they were never written down they will be applied inconsistently.

Build a seed set, then work outwards

Filters applied to a whole index return the popular and the obvious, which is selection bias built into the interface rather than into your judgment. A better first move is to name five to ten creators you already believe are right, from campaigns you have seen, from your own customers, or from the last time you did this.

Then ask what those people share. Not their follower counts. The format they use, the questions their audience asks, the adjacent categories they drift into, the times of year they get busy. Those shared properties are your real search conditions, and several of them will be things no filter in any product can express.

From the seed set, three moves widen the list.

• Audience adjacency. Who else reaches the same people, including creators in a category you would not have thought to search.
• Format adjacency. Who makes the same kind of thing for a different subject. Format travels better than topic for demonstration-led work.
• Position adjacency. Who occupies the same position in a smaller market. This is where the underpriced candidates are.

The tool matters most in the widening step, because that is where an index that reaches beyond the obvious earns its money. How those indexes are built, and what each kind structurally cannot see, is set out in the guide to creator discovery data.

The exclusion list is the strategy

Every campaign generates knowledge about who not to approach: people who declined, people whose rates put them out of reach, people with a conflict, people who were fine and simply did not suit the brief. Almost nobody keeps it.

Keep it, with a reason and a date against each name. A reason lets you reverse a decision when circumstances change, and a date stops a two-year-old refusal from blocking someone forever. Within a year it removes a large slice of work from every search, and it is portable in a way a vendor's saved search is not.

Pair it with the opposite list: people you have worked with, what was agreed, and how it went. That record is the strongest filter you will ever have, and building it depends on your own measurement being consistent enough to compare campaigns, which is the argument in the guide to analytics layers.

Knowing when to stop

Searching has no natural end, so give it one before you start, or the exercise turns into analysis paralysis with a deadline attached. Two rules work.

Stop when the last twenty results contain nobody you would contact. That is the index telling you it has been mined for this brief.

Stop when the shortlist has enough candidates to survive its own attrition. Expect refusals, unavailability and rates you cannot meet, so a list that produces the campaign you need has to be several times longer than the campaign. Work out that multiple from your own history rather than from a rule of thumb.

A shortlist is not a ranking, and treating it as one causes the classic failure: booking down a list in order until the budget runs out. Group it instead by role in the campaign, since the person who explains the product and the person who makes it look normal are doing different jobs.

Who searches, and what happens next

Searching is skilled work that looks unskilled, which is why it gets given to whoever is free. The person doing it needs the brief, the constraints and the authority to reject, and they should be the person who writes the shortlist reasons.

Where an agency does it for you, ask to see the rejections rather than the selections, because the rejections show you the standard being applied. What agencies actually do well, and where they do not, is separated in the guide to agency models.

Discovery ends at a shortlist. What follows is contracting, briefing, tracking and payment, which is a different job with its own failure modes, covered in the guide to campaign workflow. Where a shortlist is heading toward long-term partnership rather than a single campaign, the counterparty is likely to be a manager, and the terms change accordingly, as described in the guide to talent representation.

Common questions

How long should a search take?

Less time than the briefing that precedes it. If searching takes days, the conditions were not settled before it started, and you are using the tool to think.

Should we search fresh each time or reuse saved searches?

Reuse the conditions, not the results. Saved results go stale quietly, and a stale result is worse than no result because it looks current.

Can this be handed to a junior person?

The mechanics yes, the rejection standard no. Pair them: one person runs the search and drafts the list, one person applies the conditions that were written as sentences.

We keep shortlisting the same names. What is wrong?

Your conditions are describing the obvious answer, or you are searching one index that ranks by popularity. Start from a seed set of people you had to work to find, and widen from those instead.