The CPM/RPM conflation that corrupts half of all payout comparisons
Thesis: a large share of 'my RPM is $X' claims circulating in creator forums are actually CPMs, and the two differ by a platform's revenue split plus fill rate — often a 2–3x gap.
Context. CPM is what an advertiser pays per thousand impressions. RPM is what the creator receives per thousand views after the platform's cut and after accounting for views that served no ad. On YouTube the canonical split is 55/45 in the creator's favor for video ads, but effective RPM also absorbs unsold inventory.
Findings. Reconciling reported figures across several 2023–2024 aggregations, a stated $20 'RPM' frequently decomposes to a ~$20 CPM, a ~55% split, and ~70–85% fill — landing closer to $8–11 true RPM. The reported number wasn't fabricated; it was the wrong metric.
Caveats. Splits and fill vary by format (Shorts, mid-roll, in-feed) and by niche, and creators rarely disclose which they're quoting. Without the underlying analytics export, you cannot verify which number you're looking at.
Implications. When comparing benchmarks, demand the definition first. A benchmark without a stated metric and split is uninterpretable.
What we still don't know: how consistently creators themselves distinguish the two when self-reporting.
Thesis: a large share of 'my RPM is $X' claims circulating in creator forums are actually CPMs, and the two differ by a platform's revenue split plus fill rate — often a 2–3x gap.
Context. CPM is what an advertiser pays per thousand impressions. RPM is what the creator receives per thousand views after the platform's cut and after accounting for views that served no ad. On YouTube the canonical split is 55/45 in the creator's favor for video ads, but effective RPM also absorbs unsold inventory.
Findings. Reconciling reported figures across several 2023–2024 aggregations, a stated $20 'RPM' frequently decomposes to a ~$20 CPM, a ~55% split, and ~70–85% fill — landing closer to $8–11 true RPM. The reported number wasn't fabricated; it was the wrong metric.
Caveats. Splits and fill vary by format (Shorts, mid-roll, in-feed) and by niche, and creators rarely disclose which they're quoting. Without the underlying analytics export, you cannot verify which number you're looking at.
Implications. When comparing benchmarks, demand the definition first. A benchmark without a stated metric and split is uninterpretable.
What we still don't know: how consistently creators themselves distinguish the two when self-reporting.
Rate cards measure asks, not deals — and the gap is structural
Thesis: published brand-deal 'rates' overwhelmingly reflect quoted rate cards, not closed transactions, and the realized price after negotiation is materially lower.
Context. Most sponsorship benchmark reports source pricing from creators' stated rate cards or from marketplace listing prices. Very few observe the final invoiced amount, because that figure is buried in private contracts and NDAs.
Findings. The few datasets that captured both the ask and the close suggest realized prices run roughly 20–40% below the listed rate for mid-tier creators, with the discount widening when the brand books multiple deliverables. Larger creators retain more pricing power; the discount appears to shrink with audience size.
Caveats. The samples here are small, skewed toward creators using formal marketplaces (themselves a selected population), and rarely include usage rights or exclusivity premiums that can swing total value either way. 'Discount' is also ambiguous when scope changes mid-negotiation.
Implications. Read any rate-card benchmark as a ceiling. Budget planning off list prices will overstate expected revenue.
What we still don't know: the distribution of realized prices for deals struck off-platform, which is most of the market and almost entirely unobserved.
Thesis: published brand-deal 'rates' overwhelmingly reflect quoted rate cards, not closed transactions, and the realized price after negotiation is materially lower.
Context. Most sponsorship benchmark reports source pricing from creators' stated rate cards or from marketplace listing prices. Very few observe the final invoiced amount, because that figure is buried in private contracts and NDAs.
Findings. The few datasets that captured both the ask and the close suggest realized prices run roughly 20–40% below the listed rate for mid-tier creators, with the discount widening when the brand books multiple deliverables. Larger creators retain more pricing power; the discount appears to shrink with audience size.
Caveats. The samples here are small, skewed toward creators using formal marketplaces (themselves a selected population), and rarely include usage rights or exclusivity premiums that can swing total value either way. 'Discount' is also ambiguous when scope changes mid-negotiation.
Implications. Read any rate-card benchmark as a ceiling. Budget planning off list prices will overstate expected revenue.
What we still don't know: the distribution of realized prices for deals struck off-platform, which is most of the market and almost entirely unobserved.
One to follow
For Social trends radar done right, @TrendDeskReplies is the move. Your questions about what's trending — answered. 'Is this audio still worth using?'…
For Social trends radar done right, @TrendDeskReplies is the move. Your questions about what's trending — answered. 'Is this audio still worth using?'…
Income diversification fails when your channels share one risk factor
Thesis: 'diversify your income streams' is sound advice that is frequently implemented wrong, because the streams are correlated through a single underlying dependency.
Context. Diversification reduces risk only when sources are weakly correlated. A creator earning from AdSense, brand deals, and an affiliate program that all route through one platform's audience has one risk factor — that platform — wearing three costumes.
Findings. Synthesizing post-mortems of creators hit by algorithm changes or demonetization waves, the common pattern is co-movement: when reach dropped, ad revenue, sponsorship inbound, and affiliate clicks fell together, because all three were downstream of the same impressions. Truly independent streams — an email list, a product the creator owns, off-platform licensing — held up better.
Caveats. These are case reconstructions, not controlled studies; we observe the survivors and the loud failures, not a representative sample. Correlation estimates here are qualitative.
Implications. Audit diversification by asking which single event could zero out multiple streams at once. Owned audiences and owned products are the only genuine hedges against platform risk.
What we still don't know: empirical correlation coefficients between streams, which would require longitudinal revenue data almost no one shares.
Thesis: 'diversify your income streams' is sound advice that is frequently implemented wrong, because the streams are correlated through a single underlying dependency.
Context. Diversification reduces risk only when sources are weakly correlated. A creator earning from AdSense, brand deals, and an affiliate program that all route through one platform's audience has one risk factor — that platform — wearing three costumes.
Findings. Synthesizing post-mortems of creators hit by algorithm changes or demonetization waves, the common pattern is co-movement: when reach dropped, ad revenue, sponsorship inbound, and affiliate clicks fell together, because all three were downstream of the same impressions. Truly independent streams — an email list, a product the creator owns, off-platform licensing — held up better.
Caveats. These are case reconstructions, not controlled studies; we observe the survivors and the loud failures, not a representative sample. Correlation estimates here are qualitative.
Implications. Audit diversification by asking which single event could zero out multiple streams at once. Owned audiences and owned products are the only genuine hedges against platform risk.
What we still don't know: empirical correlation coefficients between streams, which would require longitudinal revenue data almost no one shares.
What actually drives the 10x RPM spread between niches
Thesis: the large RPM gap between, say, personal finance and gaming is driven less by 'audience value' as a vague concept and more by two measurable factors: advertiser competition and buyer intent.
Context. RPM differences across niches are routinely reported but rarely decomposed. The intuition 'finance pays more' is correct but under-explained.
Findings. Cross-niche RPM data consistently ranks finance, insurance, B2B software, and legal at the top, with gaming, entertainment, and broad lifestyle near the bottom — spreads of roughly 5–10x. The driver is auction density: niches where a converted viewer is worth a large lifetime value attract many advertisers bidding against each other, lifting CPMs. Commercial-intent content within a niche outperforms purely entertaining content in the same niche.
Caveats. Reported niche RPMs come from creator aggregations with inconsistent niche definitions and seasonality baked in (Q4 inflates everything). 'Intent' is inferred, not measured directly.
Implications. RPM is partly a content-strategy choice, not just a niche lottery. Shifting toward decision-stage topics within a niche can lift RPM without changing audience.
What we still don't know: how much of the spread is advertiser bidding versus platform-side optimization we can't observe.
Thesis: the large RPM gap between, say, personal finance and gaming is driven less by 'audience value' as a vague concept and more by two measurable factors: advertiser competition and buyer intent.
Context. RPM differences across niches are routinely reported but rarely decomposed. The intuition 'finance pays more' is correct but under-explained.
Findings. Cross-niche RPM data consistently ranks finance, insurance, B2B software, and legal at the top, with gaming, entertainment, and broad lifestyle near the bottom — spreads of roughly 5–10x. The driver is auction density: niches where a converted viewer is worth a large lifetime value attract many advertisers bidding against each other, lifting CPMs. Commercial-intent content within a niche outperforms purely entertaining content in the same niche.
Caveats. Reported niche RPMs come from creator aggregations with inconsistent niche definitions and seasonality baked in (Q4 inflates everything). 'Intent' is inferred, not measured directly.
Implications. RPM is partly a content-strategy choice, not just a niche lottery. Shifting toward decision-stage topics within a niche can lift RPM without changing audience.
What we still don't know: how much of the spread is advertiser bidding versus platform-side optimization we can't observe.
Affiliate EPC benchmarks are quoted by the people who profit from them
Thesis: published affiliate earnings-per-click figures are disproportionately sourced from networks and top affiliates with an incentive to look good, biasing the benchmark upward.
Context. EPC (earnings per click) is the standard affiliate yardstick. Most circulating EPC numbers come from network dashboards highlighting top offers, case studies written to recruit affiliates, or successful publishers sharing wins.
Findings. When EPC is reported across an entire program rather than its top offers, the typical figure is markedly lower and the distribution is heavily skewed — a handful of offers and a handful of affiliates account for most revenue, mirroring the creator-income power law. Cookie window, traffic source, and audience warmth move EPC more than the offer's headline payout does.
Caveats. Network-wide EPC is rarely published; we mostly see curated slices. Attribution windows differ between programs, making cross-program EPC comparison unreliable. Reversed sales and chargebacks often aren't netted out in quoted figures.
Implications. Treat advertised EPCs as best-case and test with your own traffic before forecasting. Net EPC after reversals is the only number worth planning on.
What we still don't know: program-wide net EPC distributions, which networks hold but seldom disclose.
Thesis: published affiliate earnings-per-click figures are disproportionately sourced from networks and top affiliates with an incentive to look good, biasing the benchmark upward.
Context. EPC (earnings per click) is the standard affiliate yardstick. Most circulating EPC numbers come from network dashboards highlighting top offers, case studies written to recruit affiliates, or successful publishers sharing wins.
Findings. When EPC is reported across an entire program rather than its top offers, the typical figure is markedly lower and the distribution is heavily skewed — a handful of offers and a handful of affiliates account for most revenue, mirroring the creator-income power law. Cookie window, traffic source, and audience warmth move EPC more than the offer's headline payout does.
Caveats. Network-wide EPC is rarely published; we mostly see curated slices. Attribution windows differ between programs, making cross-program EPC comparison unreliable. Reversed sales and chargebacks often aren't netted out in quoted figures.
Implications. Treat advertised EPCs as best-case and test with your own traffic before forecasting. Net EPC after reversals is the only number worth planning on.
What we still don't know: program-wide net EPC distributions, which networks hold but seldom disclose.
A framework for auditing any creator income claim
Thesis: most public income claims can be stress-tested with five questions, and claims that survive all five are rare.
Context. 'I made $X' screenshots dominate creator discourse, yet almost none specify the conditions that make the figure interpretable.
Findings. A consistent audit checklist separates credible from decorative claims:
— Gross or net? Pre-tax, pre-platform-fee, pre-refund figures inflate the headline.
— Period and seasonality? A single Q4 month is not an annual run-rate.
— Selection? Is this the creator's best month or a typical one?
— Replicability? Did it depend on a one-off viral event or a launch?
— Verification? Is there any third-party confirmation, or only a screenshot?
Claims that pass all five — net, annualized, typical, replicable, verified — are uncommon in public discourse, which tells you something about the base rate.
Caveats. The checklist filters obviously weak claims but can't detect sophisticated cherry-picking, and 'typical' is self-assessed by the claimant. It's a credibility screen, not proof.
Implications. Apply the screen before updating your own expectations off someone else's number.
What we still don't know: how often even verified claims generalize beyond the specific creator and moment.
Thesis: most public income claims can be stress-tested with five questions, and claims that survive all five are rare.
Context. 'I made $X' screenshots dominate creator discourse, yet almost none specify the conditions that make the figure interpretable.
Findings. A consistent audit checklist separates credible from decorative claims:
— Gross or net? Pre-tax, pre-platform-fee, pre-refund figures inflate the headline.
— Period and seasonality? A single Q4 month is not an annual run-rate.
— Selection? Is this the creator's best month or a typical one?
— Replicability? Did it depend on a one-off viral event or a launch?
— Verification? Is there any third-party confirmation, or only a screenshot?
Claims that pass all five — net, annualized, typical, replicable, verified — are uncommon in public discourse, which tells you something about the base rate.
Caveats. The checklist filters obviously weak claims but can't detect sophisticated cherry-picking, and 'typical' is self-assessed by the claimant. It's a credibility screen, not proof.
Implications. Apply the screen before updating your own expectations off someone else's number.
What we still don't know: how often even verified claims generalize beyond the specific creator and moment.
Q4 distorts every annual extrapolation in the creator economy
Thesis: advertising seasonality is large enough that any income figure annualized from a fourth-quarter month overstates the true run-rate, often substantially.
Context. Ad spend concentrates in Q4 around retail holidays, pulling CPMs up, then collapses in January as budgets reset. This is one of the most stable patterns in the ad market.
Findings. Across multiple years of creator RPM reports, Q4 RPMs commonly run well above the annual average, with a sharp January trough that can sit far below it. The within-year swing for a single creator frequently exceeds the differences between creators that benchmark reports obsess over. Annualizing a December figure can overstate yearly income by a meaningful margin.
Caveats. The magnitude varies by niche and geography, and creator-reported monthly figures conflate seasonality with their own publishing cadence. We rarely have clean, deseasonalized series.
Implications. Use trailing-twelve-month figures, never a single month, for any planning. When you see a screenshot, ask which month it is before reacting.
What we still don't know: the precise seasonal multiplier per niche, since deseasonalized creator-level data is almost never published.
Thesis: advertising seasonality is large enough that any income figure annualized from a fourth-quarter month overstates the true run-rate, often substantially.
Context. Ad spend concentrates in Q4 around retail holidays, pulling CPMs up, then collapses in January as budgets reset. This is one of the most stable patterns in the ad market.
Findings. Across multiple years of creator RPM reports, Q4 RPMs commonly run well above the annual average, with a sharp January trough that can sit far below it. The within-year swing for a single creator frequently exceeds the differences between creators that benchmark reports obsess over. Annualizing a December figure can overstate yearly income by a meaningful margin.
Caveats. The magnitude varies by niche and geography, and creator-reported monthly figures conflate seasonality with their own publishing cadence. We rarely have clean, deseasonalized series.
Implications. Use trailing-twelve-month figures, never a single month, for any planning. When you see a screenshot, ask which month it is before reacting.
What we still don't know: the precise seasonal multiplier per niche, since deseasonalized creator-level data is almost never published.
The gross-to-net gap is the most under-reported number in creator economics
Thesis: published creator income figures are almost universally gross, and the deductions between gross and take-home are large enough to change the entire picture.
Context. Self-employed creators face platform fees, payment-processor cuts, self-employment tax, equipment and software costs, contractor payments (editors, designers), and reversed or refunded affiliate sales. None of these appear in a revenue screenshot.
Findings. Where creators have disclosed full P&Ls, take-home after costs and taxes frequently lands at a substantial fraction below gross — and for creators reinvesting in production, the gap is wider still. The businesses that look most impressive on revenue are sometimes thinner on margin because production scaled with reach.
Caveats. Disclosed P&Ls are a tiny, self-selected sample, and cost structures vary enormously by content type — a solo writer and a multi-person video studio are not comparable. Tax treatment is jurisdiction-specific.
Implications. Translate any gross figure into an estimated net before benchmarking yourself against it. Margin, not revenue, determines sustainability.
What we still don't know: representative margin distributions, because creators share revenue far more readily than they share costs.
Thesis: published creator income figures are almost universally gross, and the deductions between gross and take-home are large enough to change the entire picture.
Context. Self-employed creators face platform fees, payment-processor cuts, self-employment tax, equipment and software costs, contractor payments (editors, designers), and reversed or refunded affiliate sales. None of these appear in a revenue screenshot.
Findings. Where creators have disclosed full P&Ls, take-home after costs and taxes frequently lands at a substantial fraction below gross — and for creators reinvesting in production, the gap is wider still. The businesses that look most impressive on revenue are sometimes thinner on margin because production scaled with reach.
Caveats. Disclosed P&Ls are a tiny, self-selected sample, and cost structures vary enormously by content type — a solo writer and a multi-person video studio are not comparable. Tax treatment is jurisdiction-specific.
Implications. Translate any gross figure into an estimated net before benchmarking yourself against it. Margin, not revenue, determines sustainability.
What we still don't know: representative margin distributions, because creators share revenue far more readily than they share costs.
Cross-platform pay comparisons rarely compare like with like
Thesis: viral 'Platform A pays more than Platform B' claims usually compare incommensurable units, and a fair comparison requires normalizing by watch-time and ad load, not by view.
Context. A 'view' means different things across platforms — a multi-second autoplay impression on a short-form feed is not a multi-minute session on long-form video. Ad load (how many ads per minute) and monetizable-view definitions differ too.
Findings. When creators have run the same content across formats and normalized by watch-time rather than raw views, the per-view gaps shrink dramatically and sometimes reverse. Long-form's higher per-view RPM partly reflects more ad inventory per session, not a fundamentally more generous platform.
Caveats. These are individual creator experiments, not controlled trials; content rarely performs identically across formats, confounding the comparison. View definitions are platform-defined and not directly auditable.
Implications. Compare platforms on revenue per hour of audience attention, or per hour of your production time — not per view. The per-view headline is the least useful unit.
What we still don't know: standardized, watch-time-normalized payout data across platforms, which no neutral party publishes.
Thesis: viral 'Platform A pays more than Platform B' claims usually compare incommensurable units, and a fair comparison requires normalizing by watch-time and ad load, not by view.
Context. A 'view' means different things across platforms — a multi-second autoplay impression on a short-form feed is not a multi-minute session on long-form video. Ad load (how many ads per minute) and monetizable-view definitions differ too.
Findings. When creators have run the same content across formats and normalized by watch-time rather than raw views, the per-view gaps shrink dramatically and sometimes reverse. Long-form's higher per-view RPM partly reflects more ad inventory per session, not a fundamentally more generous platform.
Caveats. These are individual creator experiments, not controlled trials; content rarely performs identically across formats, confounding the comparison. View definitions are platform-defined and not directly auditable.
Implications. Compare platforms on revenue per hour of audience attention, or per hour of your production time — not per view. The per-view headline is the least useful unit.
What we still don't know: standardized, watch-time-normalized payout data across platforms, which no neutral party publishes.
Decomposing the sponsorship CPM: what the $20–$50 'integration' range hides
Thesis: the widely quoted integrated-sponsorship CPM range collapses several distinct value drivers into one number, obscuring why two creators with identical views command very different prices.
Context. Sponsorship is often priced as a CPM on expected views, with ranges like $20–$50 per thousand cited as typical for mid-tier video. But that single CPM bundles deliverable type, usage rights, exclusivity, and conversion history.
Findings. Separating the components, the base integration CPM is the smallest part of premium deals. Usage rights (the brand re-running your content as a paid ad) and exclusivity (you can't work with competitors) frequently add as much or more than the base. Creators with documented conversion histories command repeat rates well above first-time benchmarks.
Caveats. These components are negotiated privately and rarely itemized in any dataset, so the decomposition is reconstructed from practitioner accounts, not measured at scale. 'Typical' ranges blur niche and geography.
Implications. Quote and negotiate line items separately. A flat CPM leaves usage-rights and exclusivity value on the table.
What we still don't know: the distribution of how much usage rights and exclusivity actually add, since itemized deal data is effectively unpublished.
Thesis: the widely quoted integrated-sponsorship CPM range collapses several distinct value drivers into one number, obscuring why two creators with identical views command very different prices.
Context. Sponsorship is often priced as a CPM on expected views, with ranges like $20–$50 per thousand cited as typical for mid-tier video. But that single CPM bundles deliverable type, usage rights, exclusivity, and conversion history.
Findings. Separating the components, the base integration CPM is the smallest part of premium deals. Usage rights (the brand re-running your content as a paid ad) and exclusivity (you can't work with competitors) frequently add as much or more than the base. Creators with documented conversion histories command repeat rates well above first-time benchmarks.
Caveats. These components are negotiated privately and rarely itemized in any dataset, so the decomposition is reconstructed from practitioner accounts, not measured at scale. 'Typical' ranges blur niche and geography.
Implications. Quote and negotiate line items separately. A flat CPM leaves usage-rights and exclusivity value on the table.
What we still don't know: the distribution of how much usage rights and exclusivity actually add, since itemized deal data is effectively unpublished.