How do you find creators who actually move product?

The short version

Stop sorting by follower count. Start sorting by prior organic affinity to your brand or your direct competitors, because a creator who already reaches for your product without being paid is the single strongest leading indicator that a paid partnership with them will convert. Pair that affinity signal with a calibrated peer comparison, strip out prior paid partnerships so you are not double-billing influence someone already bought, rank the survivors by Creator Value Score, and run one final audience-overlap check so you are not paying to reach people who already follow you. This playbook walks the full workflow inside Sentia, with the numbers you should expect to see at each stage.

What you need before you start

• A Sentia monitor with at least 90 days of crawl history for your brand. The creator engine needs a baseline before affinity scoring is meaningful.
• Three to five direct competitors you would genuinely benchmark against, not aspirational giants. If you sell a $34 serum, your cohort is other $25 to $50 serums, not a luxury house brand.
• Roughly 45 minutes for a first pass, plus another 30 if you plan to calibrate the full shortlist.

Step 1 · Define the cohort, not the creator

Before you list a single creator, define the brand cohort. Open the monitor for your brand, go to Settings · Cohort, and add three to five direct competitors. Sentia rebuilds the cohort footprint overnight, merging share-of-voice, mention velocity, and audience graphs across the combined set.

The cohort is the universe Sentia will search for affinity signals, so a sharp definition here pays off five steps later. A cohort that is too wide floods you with creators who mention big adjacent brands but have never touched your price tier. A cohort that is too narrow starves the pool.

Expected output: a cohort card showing total crawlable mentions per month across all members. For a mid-cap beauty brand, expect 40,000 to 120,000 monthly cohort mentions. If you see fewer than 5,000, your competitors are too niche and you should widen by one slot.

Step 2 · Pull the cross-platform affinity list

Inside the monitor, navigate to the Creators tab. Sentia queries creator_brand_mentions across Instagram, TikTok, and YouTube and returns every creator who has organically mentioned anyone in the cohort over the last 180 days. The default sort is by mention count; switch to Affinity Score to surface creators with deep, repeated overlap rather than one-off shoutouts. Affinity Score weights repeat mentions, unscripted product placement (the product visible in a routine video, not a dedicated ad read), and recency, so a creator who mentioned a cohort brand four times in the last 60 days outranks one who mentioned it seven times spread across two years.

The output here is your prospect pool. For a mid-cap beauty brand the pool usually lands between 200 and 800 creators. For a niche category, say refillable fragrance or clinical haircare, it can be as low as 50, and that is fine. Depth beats breadth.

Expected output: a creator table with columns for platform, follower count, mention count, Affinity Score, and last-mention date. Export it as CSV if your partnership team wants to work outside the tool.

Step 3 · Strip out the paid partnerships

Sentia tags each mention with its provenance. Mentions sourced from the Meta Ad Library, the TikTok Commercial Content Research API, YouTube's paid-promotion flags, or an explicit paid-partnership disclosure are paid by definition. Apply the Organic only filter and watch the pool shrink, typically by 30 to 50 percent, because most high-mention creators in any category are already on someone's roster.

What you want is the organic shoutout · the creator who reached for your brand or a direct competitor without being asked. That is the affinity signal that predicts performance. A creator who was paid to mention a competitor tells you they take briefs; a creator who did it for free tells you their audience overlaps with your buyer.

Expected output: the same table, filtered, with a provenance column showing every retained mention as organic. If a creator has both paid and organic mentions for the cohort, keep them; the organic mentions still count.

Step 4 · Rank by Creator Value Score

The remaining creators receive a Creator Value Score, which fuses reach, engagement rate, brand affinity, and partnership-cost efficiency into one geography-aware number, normalized 0 to 100. Sort descending. The top quartile is your shortlist, usually 30 to 60 creators.

Two practical notes on reading CVS. First, the geography weighting matters for fulfillment: a creator with a CVS of 78 concentrated in a market you cannot ship to is worth less to you than a 71 concentrated where you sell. Check the geography split in the creator drawer before you get attached to a score. Second, CVS is recalculated weekly, so scores you exported last month are stale; re-export before outreach.

Expected output: a ranked shortlist with CVS, estimated partnership cost band, and primary geography for each creator.

Step 5 · Calibrate against the cohort, not the world

Open the Calibrate panel on each shortlisted creator. A 4 percent engagement rate is excellent for a 2M-follower beauty creator and unremarkable for a 50k-follower one, because engagement decays with scale. The calibrated badge tells you whether the creator outperforms peers at their own size, platform, and category, which is the only comparison that is fair.

Calibration also flags anomalies: bought followers, engagement pods, and sudden engagement spikes uncorrelated with content quality. Drop creators flagged red, no matter how good their raw numbers look. Expect to lose 20 to 40 percent of the shortlist here, which is the step doing the most quiet work.

Expected output: each creator carries a green, amber, or red calibration badge plus a one-line reason when amber or red. Amber means proceed with a smaller test budget; red means cut.

Step 6 · Cross-check audience overlap

Last filter. Sentia compares the creator's follower base against your brand's known organic audience, the people who already engage with your owned content and site. Heavy overlap means you are paying to reach people who already follow you; as a rule of thumb, overlap above 45 percent means most of your spend is redundant. Light overlap with strong affinity is the sweet spot · the creator's audience expands your reach while their demonstrated product taste means they are still your buyer.

Aim for the 10 to 35 percent overlap band. Below 10 percent, treat it as a warning sign too: it can mean the affinity was a fluke or the audience is adjacent rather than complementary.

Expected output: an overlap percentage and a small Venn-style breakdown in the creator drawer, showing shared audience, creator-only, and brand-only segments.

What this gets you

For a mid-cap beauty brand running this workflow in Q2 2026, the funnel looked like this: 600 cohort creators with any mention history, 380 after the organic-only filter, 47 with prior organic affinity strong enough to rank, 19 who survived calibration, and 9 who cleared the overlap band. Nine names, each with documented prior affinity, a defensible CVS, clean calibration, and overlap math that says they grow reach rather than recycle it. That list goes to the partnership team with a confidence level no cold agency roster can match, because every name already demonstrated the behavior you are about to pay for.

Troubleshooting

The prospect pool is under 50 creators. Your cohort is too narrow or your category is too new. Add one adjacent-but-credible competitor, or extend the lookback window from 180 to 365 days in the Creators tab settings. Do not loosen the organic-only filter to compensate; paid mentions will not predict performance.

Every shortlisted creator has heavy audience overlap. Your brand has grown to dominate its niche. Shift strategy: either accept overlap above 45 percent for conversion-oriented campaigns where retargeting your own audience is the point, or add a competitor with a younger audience to the cohort and re-run Steps 2 through 6.

Scores changed dramatically between exports. CVS recalculates weekly and Affinity Score decays with recency, so a creator who went quiet for a month will slide. Re-export within 48 hours of outreach so your negotiation numbers are current.

A high-CVS creator is flagged red on calibration. Trust the flag. Bought followers and pod engagement are exactly the failure mode follower-count ranking was always vulnerable to, and calibration exists to catch what raw numbers hide. Move down the list.

The partnership team wants the raw data. Export the shortlist CSV from the Creators tab; it includes provenance, calibration status, and overlap, so the context travels with the names.

Related

• Glossary: Creator affinity, CVS · Creator Value Score, EMV, Calibration
• Comparison: Sentia vs Brandwatch