Is Substack's 10% Worth It? Discovery, Break-Even Math, and Migration Boundaries

TL;DRSubstack charges 10% of all paid-subscription revenue in exchange for publishing, payments, a recommendation network, and a shorter checkout path. Break-even depends on retained incremental value—not on subtracting 10% from the company's self-reported 25–30% network share.

#substack #creator-economy #content-business #newsletter #subscription #platform

Series: Who Controls the Creator-Reader Relationship (2 / 7)

What Platform Distribution Costs: Medium, Vocus, and the Readers You Cannot Take With You How Patreon Turns Supporters Into Members: Free Entry, Paid Tiers, and the Cost of Leaving

🌏 中文版

Imagine a market that brings customers to its vendors. Instead of fixed rent, it takes 10% from every bowl of noodles—including sales to regulars the vendor brought from home. The market's foot traffic, membership card, and saved payment details can make buying easier. Whether the fee is worthwhile still depends on how many additional customers stay, not on how crowded the stall looks today.

Substack is that kind of market. Creators can publish and send newsletters for free. Once they enable paid subscriptions, the platform takes 10% of subscription revenue. The fee buys more than hosting: payments, subscription management, recommendations between writers, App and Notes distribution, and a shorter checkout path for readers with stored payment details.

The useful question is not whether 10% is high or low. It is how much retained revenue those capabilities add to a specific publication and which alternative costs they replace. Two publishers can reach opposite answers.

The 10% buys a conversion chain

Substack's payment setup documentation states the 10% platform fee, and Axios independently reported the same rate in 2025. Stripe payment and billing fees are separate; purchases inside the iOS app may also incur an Apple fee. “The creator keeps 90%” is therefore incomplete accounting.

The product is designed to create this flywheel:

flowchart TD
    A[Recommendations<br/>Notes, App, other writers] --> B[Reader discovers a publication]
    B --> C[One-click free subscription]
    C --> D[Repeated contact by email and App]
    D --> E{Will the reader pay?}
    E -- Not yet --> D
    E -- Yes --> F[Short checkout with a stored card]
    F --> G[Publisher subscription revenue]
    G --> H[Substack takes 10% of paid-subscription revenue]

This diagram describes a product mechanism, not a proven causal effect for every writer. Recommendations, the App, Notes, guest posts, and mentions may introduce readers. Email and paywalls create repeated conversion opportunities. Stripe Connect handles payments, verification, and payouts. A publisher's own acquisition and retention data must show which steps create incremental revenue.

The 25–30% figure is company-reported, with conflicting definitions

Substack has published figures for paid subscriptions attributed to its network. They should not be drawn as one clean growth line.

Date	Public claim	Actual definition	What it supports
2022	About 10% of paid subscriptions came from the network	Entire network; earlier product	A historical company-reported baseline
2024	Recommendations drove 25% of new paid subscriptions	New paid subscriptions; one feature	Recommendations had become important, not independently audited
2025	The network drove 30% of paid subscriptions	Co-founder interview; cohort undisclosed	Latest company claim, not transferable to one writer
Current growth page	The network drives 25% of paid subscriptions	Measurement period undisclosed	May differ from 2025 because of version or denominator

Substack's 2022 post, its current growth page, TechCrunch's 2024 report, and a 2025 Hollywood Reporter interview use different features, periods, and denominators. The media reports establish when the claims were made, but the numbers still originate with the company or a co-founder rather than an independent audit.

The defensible conclusion is narrower: Substack reports that its network produces roughly one-quarter to three-tenths of paid subscriptions. Public evidence does not reconstruct the measurement or establish the same increment for an individual publisher, particularly one writing in Traditional Chinese.

You cannot subtract 10% from 25%

“The network drives 25% of paid subscriptions” does not imply a 15% net benefit after the platform fee. Network share measures subscriber source; the platform fee applies to all paid revenue. Prices, retention, refunds, and readers who would have converted anyway all change the answer.

Define four variables:

R0: annual subscription revenue available without Substack.
ΔR: incremental annual revenue created by Substack's network, payment conversion, and operating convenience.
C_alt: annual cost of the alternative platform.
C_ops: additional annual email, payment, support, legal, and development cost outside Substack.

Before differences in Stripe, Apple, taxes, and refunds, the model for the net benefit of staying is:

ΔR - 0.10 × (R0 + ΔR) + C_alt + C_ops

The break-even condition is:

0.90 × ΔR > 0.10 × R0 - C_alt - C_ops

Scenario	`R0`	`ΔR`	10% platform fee	Result before alternative and operating cost
Starting from zero; increment is the main revenue	0	10,000	1,000	+9,000
Existing revenue plus one new cohort	50,000	10,000	6,000	+4,000
Mature list with the same new cohort	100,000	10,000	11,000	-1,000
High revenue without proportional increment	500,000	25,000	52,500	-27,500

The amounts are unit-based scenarios for the formula, not dollar forecasts or promises of Substack revenue. Adding C_alt and C_ops can turn a negative result positive. Poor retention can erase a positive result.

flowchart TD
    A[Measure sources for at least 90 days] --> B{Can you identify retained<br/>paid increment from Substack?}
    B -- No --> C[Add source cohorts<br/>and free-to-paid tracking]
    B -- Yes --> D{Does incremental profit exceed<br/>10% of paid-subscription revenue?}
    D -- Yes --> E[Stay and recalculate periodically]
    D -- No --> F{Do alternative and operating costs<br/>cover the gap?}
    F -- Yes --> E
    F -- No --> G[Test Ghost or another SaaS]

Exportable email is not an exportable business

The Substack subscriber dashboard can export all subscribers or a filtered segment as CSV, using all fields or the currently displayed columns. Posts and related publication data also have an official export. Content and email are therefore relatively portable, but formatting, URLs, comments, Notes relationships, and platform recommendations do not move intact.

Paid subscriptions have a more conditional answer. The common claim that every reader must re-enter a card is too absolute. Ghost's Substack migration guide says paid memberships can move when conditions are met and the same Stripe account is used. Platformer's migration announcement likewise said existing subscribers did not need to act.

“The same Stripe account” does not mean zero friction. Ghost's documentation warns that Substack's 10% fee on old subscriptions may require separate coordination to remove. If a publisher directly disables paid subscriptions in Substack, Substack says existing subscriptions are canceled and prorated refunds are issued. The accurate boundary is that payment migration depends on the Stripe account, source platform, destination, and support process. It requires more than an email CSV, but it does not always require every reader to re-enter a card.

flowchart TD
    A[Prepare to leave Substack] --> B[Export content and subscriber CSV]
    B --> C{Can the same Stripe account be used?}
    C -- Yes --> D[Reconcile customers, subscriptions,<br/>and fee handling with the destination]
    C -- No or uncertain --> E[Plan complimentary access,<br/>resubscription, and notices]
    D --> F[Test access, renewal, and refunds]
    E --> F
    F --> G[Rebuild URLs, recommendations,<br/>and Notes acquisition]

Platformer shows that two things can be true

Casey Newton announced Platformer's departure from Substack in 2024. His first-party post said the publication's free audience had grown substantially in the prior year and attributed some of that growth to Substack's recommendation and network tools. Because the exact figure lacks a second fully readable independent source, this article keeps only the narrower conclusion: platform distribution provided real value to Platformer.

The same publication still moved to Ghost. As an established publication's existing revenue grows, a 10% cost grows with it; brand, content governance, and control may also receive more weight. This does not prove that every mature writer should leave. It shows that one platform can produce different answers at the launch and mature stages.

The largest unknown for Traditional Chinese publishers is network density

Substack's recommendation flywheel needs readers already on the platform, adjacent writers willing to recommend one another, and enough same-language content to continue exploring. This research found no network-share cohort for Traditional Chinese or Taiwan, so the platform-wide 25–30% cannot be transferred to that market.

Taiwan-based creators must also verify Stripe availability, cross-border card success, refunds, taxes, and invoicing. Multi-currency payment support does not establish that Substack handles every Taiwan-specific obligation.

The smallest useful test runs for 90 days. Separate direct, Google, social, Substack App, and Recommendations acquisition. For each source, record free-to-paid conversion, 90-day retention, and refunds. If the dashboard cannot answer, add campaigns, UTMs, or staged invitations. Without a creator's own cohort, there is no creator-specific answer to the 10% question.

AI changes the questions, not the proven answer

Generative AI lowers the cost of supplying content and may raise quality and trust costs for a recommendation network. A WIRED sample of popular Substack publications found that some used AI, but detectors can be wrong and the sample cannot estimate a platform-wide share. Substack's Content Guidelines address plagiarism, spam, phishing, and inauthentic activity; they do not remove all AI-generated content solely because AI was used.

AI search may also intercept some open-web clicks. In a 2025 Pew sample of U.S. Google users, traditional-result clicks were lower when an AI summary appeared. That supports the direction of building a direct email relationship. It does not prove that AI has raised Substack paid conversion or made the 10% fee more valuable.

The final decision is straightforward: treat the 10% as an acquisition and operations expense that scales with paid-subscription revenue, then recalculate it each quarter using incremental revenue, retention, and replacement costs. Substack's network can be a real product advantage. Until a publisher's own data appears, it is not a guaranteed return.