A supplement brand asks for fitness creators. The shortlist comes back with a powerlifter, a yoga teacher, a marathon runner and a home-workout account. Every one of them is a fitness creator. No two of them share an audience in any meaningful way, and at most one of them is right for the product.
Nobody made a mistake. The filter did exactly what it was asked to do. The label was the problem, and the label is where most shortlists go wrong before a single profile is opened.
What "fitness" actually contains
At least five distinct audiences, with different purchase intent and only partial overlap.
- Strength training. Committed, technical, long time horizon. Buys equipment, programming and coaching. The comments argue about form.
- Home workouts and bodyweight. Time-poor, beginner to intermediate, buys convenience and apps. Shares a vocabulary with strength training and almost nothing else.
- Running and endurance. Buys shoes, race entries, wearables and performance nutrition. Close to a separate sport with a separate media culture.
- Yoga, pilates and mobility. Overlaps with wellness and mental health far more than with anything in a gym.
- Physique and aesthetics. Shares its subject with strength training and almost none of its intent. The audience is there for the body, not the method.
Nutrition sits across all of them and belongs to none: recipe accounts, macro-tracking accounts and supplement-review accounts each pull a different crowd.
Sponsorship fit does not follow the label. It follows purchase intent, and purchase intent lives in the sub-niche. A protein brand and a yoga-mat brand are both "fitness" advertisers with almost no overlap in who they should be paying.
Beauty, travel, and the special case of lifestyle
Beauty splits at least into skincare-as-routine, skincare-as-ingredient-science, makeup artistry, everyday get-ready-with-me, hair, nails and fragrance. An ingredients audience is sceptical and reads labels. A GRWM audience is buying a mood. Pitch the wrong one and the same product lands very differently.
Travel splits into budget backpacking, luxury, family, digital nomad, adventure and city guides. A hostel chain and a five-star resort both search "travel creators" and should never get the same list.
Lifestyle is the worst offender, because it is two completely different things wearing one word. Sometimes it is the label a creator picks when their content has no subject: a genuine absence of a niche. Sometimes it is a creator whose audience follows them across several real subjects, which is valuable, rare, and very hard to brief.
Those two cases have opposite implications for a campaign, and no filter can tell them apart from the label alone. When "lifestyle" appears on a shortlist, treat it as a flag to go and read the content, never as a category.
The benchmarking problem
This is the expensive consequence, and it is subtler than the shortlist one.
If your engagement median for "fitness" is computed across all those sub-niches, it is an average of populations with genuinely different baselines. Comparing an individual creator to it produces a number that is arithmetically correct and analytically useless.
It misleads in both directions at once. A creator in a sub-niche that typically engages hard looks unremarkable against the blended median and gets cut. A creator in a sub-niche that typically engages lightly looks strong against the same median and gets bought. You have not measured quality; you have measured which sub-niche each creator happens to be in.
The number is not wrong. It is answering a question you did not ask. What counts as a good engagement rate is almost entirely a question about which peer group produced the comparison.
There is a real tension here, and it is worth stating rather than hiding. The tighter the peer group, the more meaningful the comparison and the smaller the sample behind it. A median drawn from thirty creators in a narrow sub-niche is noisier than one drawn from three thousand across a category. There is no setting that gives you both, and any tool implying otherwise is overselling.
Finding the real sub-niche from the content
Six steps, ten minutes per creator, and you only run it on a shortlist.
- Ignore the bio. Bios are positioning. They are written once, often aspirational, frequently stale. "Fitness | Nutrition | Mindset" describes nobody.
- Label the subject of the last thirty to fifty posts. The subject, not the format. What is the post actually about?
- Sort by proportion, not presence. One yoga post does not make a yoga creator. You are looking for what dominates, and for whether anything does.
- Read the comments for vocabulary. This is the fastest single check. A strength audience talks about load, cues and PRs. A physique audience talks about the body. A running audience talks about pace and distance. The audience declares its sub-niche even when the creator does not.
- Check who already sponsors them. Existing partnerships are other buyers' conclusions about this audience, arrived at with budget on the line.
- Check the performance split, not just the posting mix. A creator who posts mostly recipes but whose training content beats their own baseline has an audience their posting mix misrepresents. Ranking their themes against their own median is what separates what they make from what works.
Steps 2 and 3 are also the cheapest defence against the lifestyle problem: if no subject accounts for a meaningful share, you have found a creator whose audience is following a person, and you should brief accordingly or not at all.
What to write in the brief instead
Replace "fitness creators, 50k+" with something a colleague could check without asking you what you meant.
Strength-training creators whose posted content is at least half barbell or gym-floor training, audience at least 60% in Romania, 20,000 to 150,000 followers, engagement at or above the median for strength training specifically at that follower band. Yoga, running and physique-first accounts are out of scope for this brief.
The last sentence is the one that stops the list drifting. Writing the exclusions down converts a vague preference into a filter, which is step one of the research workflow and the step that most determines whether the rest of it is any good.
Where this gets you into trouble
Sub-niches are a tool for thinking, not a fact about the world.
- There is no canonical taxonomy. Reasonable people split "fitness" differently, and any tool's categories (including Lyren's) are choices somebody made. Ask what the categories are before you trust a category median.
- Over-narrowing is a real failure. Slice thin enough and you have four candidates and no benchmark to judge them against. The correct grain is the finest one where enough creators remain to compare.
- Creators move. An account can shift sub-niche across a year, and a twelve-month labelling window averages the transition into something describing neither end of it.
- Adjacency sometimes works, and the point is that it should be chosen. A running creator can sell a protein product. The rule is not that crossover fails; it is that crossover should be a decision with a reason, not a side effect of a broad filter.
- Some creators genuinely are multi-subject. The method above will label them badly, and they are often the most interesting names on a list.
- Thin samples make noisy medians. In a narrow sub-niche in a small market, the benchmark may rest on very few creators. That should be visible to you, not smoothed into a confident number.
Where Lyren fits
Lyren's authority score is computed within the niche rather than across a platform, which is the whole reason the niche definition matters as much as it does. A score that blended strength training with yoga would be measuring the wrong thing precisely, which is worse than measuring it roughly.
Coverage is deliberately narrower than the largest indexes, market by market, because a benchmark is only worth having where there are enough comparable creators behind it. Where the sample in a niche is thin, the honest answer is that it is thin. Access is invite-only via the waitlist, starting with Romania.