How IU scores
Pricing Benchmark, Reach Verification,
and Regulatory Compliance

Every creator on the index gets three USD figures — low, mid, high — for any platform / format / exclusivity combination. They come from one transparent formula:

base   = (views / 1000) × format_mult × (1 + exclusivity_add)

low    = base × cpm_low
mid    = base × cpm_anchor
high   = base × cpm_high

views is the creator's average recent non-Shorts views, captured from the YouTube Data API. cpm is the published per-niche band (section 03). format_mult adjusts for placement type (section 04). Exclusivity is an optional, separately surfaced uplift (section 05) — never folded silently into base.

The mid figure is the band's anchor (the median). Low and high are the band's actual edges from industry calibration — not arbitrary ±50% bounds.

This is the single most important design choice in the model. The legacy approach across the influencer-marketing industry prices by subscriber count. We think that is fundamentally wrong, and the data is overwhelming.

Brands buy attention, not follower lists. A paid placement reaches the viewers who watch the video it sits inside, not the historical population that once hit subscribe. Subscriber counts decay over time: there is no un-subscribe signal in the YouTube API. Once a viewer counts, they always count, even if they haven't opened the app in three years.

Average recent views are the right unit because:

• They reflect what the creator actually delivers on a sponsored placement today.
• They auto-adjust for platform-algorithm shifts and audience drift.
• They surface “zombie” channels — large historical subscriber bases that no longer convert to impressions.

In our own US-equities calibration (sample of mega-tier channels captured May 2026), the median mega creator (1M+ subscribers) gets ~3.4% of subscribers as average views per upload; about half of mega-tier finance channels sit below that. Two creators with identical 1M-subscriber counts can have a 40× gap in actual reach. Our view-based model prices them 40× apart automatically. A subscriber-based model would price them identically.

CPM (cost per mille) is the niche-level lever for brand spend per impression. We publish a transparent low / anchor / high band for each of the 26 niches in the index, rather than a single point estimate. The band width reflects deal-to-deal variance even within a tightly-defined niche.

Finance is the highest band on the platform. Published 2024-2025 industry surveys consistently put financial-services video CPM in the $50-80 range, the highest of any consumer-facing vertical. The four highest bands — finance, crypto, b2b, business — are all priced from the same underlying truth: financial-services advertisers pay the most because the customer lifetime value justifies it.

Channels flagged us_equities_coverage = true get the finance band even if their primary niche classifier put them in business — the override exists so US-equities creators always price at the IR-relevant rate. (US-equities band: $45–$75, anchor $55.)

Default fallback band when a creator is unclassified: $9–$16 (anchor $12).

Multiplier applied to (views / 1000) × cpm to account for production cost and brand real-estate of each deliverable. The index is video_integrated at 1.0×.

Brand-side exclusivity (a non-compete window during which the creator cannot place a competitor) is a separate transparent line item, never folded silently into the base rate. Brands see exactly what they pay for; creators can negotiate the window independently of the integration price.

We do not yet have an avg_views_recent capture for every creator on the index — the metric is rolling out niche by niche, with US-equities currently complete and the broader long tail in progress.

When average views are missing, the engine falls back to a tier-based estimate:

views = subscribers × per_tier_median_ratio

nano   (10K–50K subs):  0.158
micro  (50K–100K subs): 0.172
mid    (100K–500K subs):0.186
macro  (500K–1M subs):  0.121
mega   (1M+ subs):      0.034

Per-tier ratios are calibrated from the actual US-equities sample. Estimates that fall back to this path are tagged confidence: low in the API and labelled accordingly in the UI; the moment avg_views_recent is captured the estimate switches automatically to confidence: high.

We deliberately avoided a single global ratio. Engagement decays non-monotonically with audience size — peaking around mid-tier (100K-500K subs) and collapsing at mega-scale (the “zombie” distribution above). One global ratio would systematically over-price megas and under-price the small-but-engaged tier that's often the highest-converting placement.

The price (Layer 1) is one thing. The influence profile (Layer 2) is what we publish alongside the price, never inside it. It is a transparency surface — a buyer can compare two creators side by side and see why one's avg_views are higher than the other's, without that judgement being smuggled into the rate.

Profile fields:

• Size tier — nano (10K-50K), micro (50K-100K), mid (100K-500K), macro (500K-1M), mega (1M+). A Postgres generated column from subscriber_count so the bucket can never drift out of sync.
• Activity label — derived from the creator's views_to_subs_ratio (avg_views ÷ subscribers). Thresholds (tier-agnostic, calibrated from the US-equities sample): low activity < 0.01 · below average 0.01-0.05 · healthy 0.05-0.50 · highly active ≥ 0.50.
• Coverage signals — primary niche, country, language, and the us_equities_coverage flag.

The activity label never enters the price. A low-activity channel is already priced down in our model because its avg_views are already low. Applying an additional engagement-penalty multiplier on top would be double-counting. We do not hide a penalty inside a price input.

Finance is the only major creator vertical regulated by an explicit U.S. federal anti-touting statute. Section 17(b) of the Securities Act of 1933 requires anyone receiving consideration to promote a security to disclose, in the same medium and at the same prominence as the promotion itself, the source, amount, and form of the compensation.

This is the legal layer that separates finfluencers from every other paid-placement market. A non-disclosed paid stock pitch is not a community-guidelines issue — it is a federal-securities violation. The clearest §17(b) precedent in the social-media era is the October 2022 Kim Kardashian / EthereumMax settlement, in which the SEC charged her for promoting the EthereumMax token on social media without disclosing she had been paid for the post; she settled the §17(b) anti-touting charges with the Commission.

Section 17(b) is a disclosure statute and is distinct from the SEC's broader securities-fraud enforcement against finfluencers. The 2022 Atlas Trading complaint — against eight Twitter / Discord finfluencers for a coordinated pump-and-dump — is a securities-fraud and wire-fraud case, not a §17(b) case; it's cited here only to signal that the SEC pursues finfluencer-marketing violations across multiple statutes, with criminal exposure attached.

We track disclosure status as an explicit public profile dimension. Every creator carries one of:

• not_assessed — we have not yet inspected this creator's content for §17(b) disclosure language. Not a clean bill of health.
• disclosed — observed prominent “#ad” / “sponsored by” language on relevant promotional content.
• undisclosed — observed promotional content that does not appear to carry disclosure.
• unknown — assessed but cannot determine with confidence.

Today every creator is not_assessed; the content-scanning infrastructure is rolling out. The slot exists in the public profile schema now so that IR / issuer clients can begin filtering on it as the data populates.

The index is built bottom-up from public data on the YouTube Data API v3. We do not buy third-party creator lists; every channel was surfaced by one of four primitives:

• Top-channel seed lists from Wikipedia (most-subscribed YouTube channels, regional and per-language variants).
• Trending sweeps via videos.list?chart=mostPopular across ~50 ISO region codes.
• Niche-targeted search via search.list?type=channel with regionCode + relevanceLanguage biases. ~300 finance-vocabulary queries across 16 languages, with deliberate APAC and emerging-market depth.
• Comment-thread harvesting as a supplemental long-tail source.

Each channel is enriched with public metadata only (title, description, country, default language, subscriber count, view count, video count). We never request OAuth scopes and never see private user data.

We also explicitly classify channel entity_type (creator / media / brokerage / institution / unknown) so that news-media outlets, exchanges, brokerages, and central banks are surfaced and labelled but filtered out of the finfluencer marketplace surface — they will never take paid placements, and conflating them with paid-promotable creators would corrupt the index.

The methodology synthesizes four categories of external grounding. We name the categories honestly rather than fabricating point citations:

• Academic research on financial influencers (“finfluencers”). A growing body of 2022-2024 peer-reviewed working papers from finance academia (Swiss Finance Institute and similar venues) documents finfluencer market structure, follower-return correlation, and the systematic disclosure gaps that motivate Layer 3. We use the directional findings to frame the activity-threshold bands and the compliance slot — not specific point estimates.
• Published industry CPM benchmarks. Influencer Marketing Hub annual reports, IZEA Insights, Tubular Labs, and creator-economy press from The Information, Bloomberg, and The Verge. Where surveys disagree, we use the median. The finance band ($45-75, anchor $55) sits inside the consensus financial-services video range.
• U.S. Securities and Exchange Commission rules. Section 17(b) of the Securities Act of 1933, public SEC enforcement actions against undisclosed paid promotion (the 2022 Kim Kardashian / EthereumMax settlement) and against broader finfluencer securities fraud (the 2022 Atlas Trading complaint, a distinct fraud track rather than a §17(b) case), and SEC investor-alert bulletins on social-media stock promotions. These ground the Layer 3 disclosure framework.
• InfluencerUnion first-party data. The index is a continuously updated dataset of ~5,300 working YouTube creators, classified by niche and entity_type, with the US-equities subset captured for avg_views_recent (validated May 2026). Spans 50+ countries and 16+ languages, with deliberate depth in Mandarin (zh-CN / zh-TW / zh-HK / overseas-Chinese diaspora), Japanese, and Korean — markets where US-equities retail interest is concentrated and other indices systematically under-cover.

Every constant in the formulas above is exported from @influencerunion/pricing-engine in this repository. If you think one is off, the engine is open and the discussion is welcome.

• Estimates, not contracts. Every dollar figure on this site is an algorithmic estimate. Real negotiated deals vary substantially on usage rights, asset licensing, brand history, content quality requirements, and creator availability.
• Estimates are benchmarks, not creator-published rate cards. Creators can claim their profile and submit corrections or self-reported rates that flow through the calibration layer.
• Disclosure compliance is “not_assessed” for all creators today. Do not read the absence of a finding as compliance with §17(b). Content-scanning rolls out platform-by-platform.
• YouTube-first coverage. The current index is YouTube-only. TikTok, Instagram, X, and Bilibili are supported in the pricing model and schema, but ingestion workers for those platforms ship in later phases.
• Average views = recent-uploads mean, not retention-weighted. We do not yet adjust for sponsored-vs-organic view-through rate or for viral-clip retention. This tends to overstate prices for creators whose viral hits dominate their recent-uploads mean.
• Recency of CPM calibration. Bands are calibrated to publicly-published industry data and our own first-party sample; they lag the market by 6-12 months. We do not have access to private rate cards.
• Language coverage skew. Discovery runs in 16+ languages with deliberate APAC depth. Bengali, Tamil, Telugu, Egyptian Arabic, Swahili, and most Sub-Saharan African languages remain under-covered relative to their actual finfluencer footprint. Expanding these is on the roadmap.