Ranking email platforms by monthly price ranks the wrong thing. Sending is a commodity and has been for years. The bill is the smallest of the three costs a platform imposes, and it is the only one anyone compares.
The other two are what this page ranks: what the platform takes from what you earn, and what it keeps when you leave.
The short answer
A revenue share is the most expensive pricing model that exists for a publication that succeeds, and the cheapest for one that does not. A flat bracket fee is the reverse. Which is better for you depends entirely on a number you do not know yet, which is why the exit cost matters more than either: it determines whether you can change your mind.
Rank platforms by whether you can leave, then by the share, then by the bill. Almost everyone ranks them in the opposite order, and the bill is the one figure that stops mattering the moment the publication works.
The three costs
The bill. Real, visible, and the least interesting. Every platform here prices in brackets by list size, so the entry rate tells you almost nothing about what you will pay at ten thousand subscribers.
The share. A percentage of subscription revenue is not a fee, it is a partnership you did not negotiate and cannot exit without a migration. It scales with your success, which is the opposite of how infrastructure should behave.
The exit cost. This is the one nobody checks before signing up and everybody discovers while leaving. A list you can export is not the same as a list you can move. An archive you can download is not the same as an archive you can rehost. The gap between those pairs is the real lock-in, and it does not appear on any pricing page.
The ranking
Ordered by total cost to the loop, cheapest first. The bill is the entry rate only.
| Platform | Entry bill | Share of paid subs | What the exit costs | Dimension it caps |
|---|---|---|---|---|
| Self-hosted sender | Server plus per-send | 0% | Nothing. You hold the database | None. Caps your time instead |
| Ghost(Pro) | From $15/mo yearly | 0% | Low. Full export, portable archive, own domain | None structural |
| Kit | Free to 10,000 | 0% on subs, fee on product sales | Low. Full list export, own sending domain | None structural |
| Buttondown | From about $9/mo | 0% | Low. Plain export, own domain | Monetization design, by omission |
| MailerLite | From about $12/mo | 0% | Low, but no publishing layer to take | Monetization design |
| beehiiv | Free to 2,500 | 0% on the paid tiers | Moderate. Monetization is gated, not owned | Monetization design on the free tier |
| Substack | $0 | 10%, indefinitely | High. See below | Owned hub strength, Distribution loop |
Entry rates read from vendor pricing pages on July 28, 2026. Every vendor here prices by list size, several have changed their free tiers in the last twelve months, and two have restructured their plans entirely. Treat these as floors and read the current page before budgeting.
How the pricing models actually behave
Four shapes cover every platform here, and each is cheapest in a different place. The reason vendor comparisons are so unhelpful is that they compare entry rates across models that do not have comparable curves.
| Model | Scales with | Cheapest when | Most expensive when |
|---|---|---|---|
| Revenue share | What you earn | You earn nothing | You earn well |
| Subscriber brackets | List size, paying or not | Small list, high revenue per subscriber | Large free list, sponsorship model |
| Send volume | Messages dispatched | Infrequent sends | Daily publishing |
| Flat plus infrastructure | Almost nothing | Any scale | Never, in money. Constantly, in time |
The subscriber-bracket trap deserves stating explicitly because it is the one that ambushes a growing publication: a free subscriber usually costs the same as a paying one. Eight hundred free signups and two hundred members is a thousand subscribers on the bill. If the model is a large free audience monetised by sponsorship rather than a small paid one, bracket pricing is working directly against the strategy.
Where a revenue share crosses over
The arithmetic is simple and almost nobody runs it before committing. Taking the one share on this table that is definitionally known, 10 percent, against an illustrative flat platform cost of $50 per month:
| Annual subscription revenue | Cost at 10% share | Cost at $50/mo flat | Which is cheaper |
|---|---|---|---|
| $3,000 | $300 | $600 | Share |
| $6,000 | $600 | $600 | Break-even |
| $25,000 | $2,500 | $600 | Flat, by $1,900 |
| $100,000 | $10,000 | $600 | Flat, by $9,400 |
| $250,000 | $25,000 | $600 | Flat, by $24,400 |
The flat figure is an illustration with a stated assumption, not a quoted price, and a real flat plan would step up with list size rather than holding at one rate. The shape is the point. A share is a rounding error below the crossover and a salary above it, and the crossover on a modest paid publication arrives in the first year or two rather than at scale.
What makes this worse than a simple pricing mistake is that the share is the one cost you cannot renegotiate by downgrading a plan. Escaping it requires a migration, and migrations get harder as the archive grows, so the cost rises at exactly the rate the incentive to leave does.
Why Substack is last despite being free
Because the bill is the smallest cost and Substack is expensive on the other two.
The share is permanent. Ten percent of paid subscription revenue, for as long as those subscriptions exist. It does not step down at scale. At meaningful revenue it exceeds every flat fee on this table by an order of magnitude, and it continues after you migrate unless you specifically get it removed.
The export is incomplete in ways the marketing does not mention. The zip contains posts as CSV and HTML, the email list, and stats. It does not contain your images: the HTML references Substack’s CDN, so the archive looks intact locally while being served by the platform you left. Substack’s own documentation states that exporting subscriber names is not currently possible. Notes, comment threads and drafts are not included either.
The addresses may not be yours. On a custom domain, they are. On a substack.com subdomain, every URL you have ever accumulated a link to belongs to the platform. That is not an export problem, it is an ownership problem, and no export format fixes it. The mechanics of moving anyway are in the Substack to Ghost migration entry.
None of this makes Substack a bad place to start. It makes it an expensive place to have succeeded, which is a different claim, and the crossover arrives sooner than most publishers plan for.
The exit cost, itemised
Exit cost is treated as one thing and is actually six, only two of which appear in any vendor’s export documentation. This is the checklist a teardown works through, because every item is verifiable from outside.
| Component | Question | Recoverable later? |
|---|---|---|
| The addresses | Can the list be exported in full, with status and metadata? | Yes, while you still have access |
| The media | Are images in the export, or referenced from the vendor’s CDN? | Yes, with work |
| The URLs | Is the archive on a domain you registered? | No |
| The billing | Is the payment processor account yours? | Only by asking members to re-enter cards |
| The sending identity | Do SPF and DKIM align with a domain you control? | Yes, and the reputation restarts |
| The history | Do open, click and signup-source records export? | No |
Two rows are marked as unrecoverable and they are the two nobody checks. The domain question is settled on the first day of the publication by whoever registered a domain, or did not. The engagement history question is settled by the platform’s export format, and losing it means every segment, every re-engagement rule and every judgement about which subscribers are real restarts from zero on the day of the move.
The sending identity is a cost, not a feature
A platform that sends from its own domain on your behalf is holding your sender reputation. It works, it is usually more deliverable than a cold domain, and it is not transferable. When you leave, the reputation stays with the vendor and your new domain starts from nothing with a list that has no history against it.
This is why the first send after any migration is the riskiest one a publication makes, and why a platform that supports authentication on your own domain from the beginning is cheaper on exit even where it is more expensive on the bill. The requirements, and how to verify alignment rather than trusting a dashboard, are on the email authentication reference page.
The gated-monetization pattern
beehiiv and Ghost both do a version of this, and it is worth naming because it distorts the comparison.
A generous free tier that excludes monetization is not a free tier for a publication that intends to earn. beehiiv’s free plan runs to 2,500 subscribers and includes no paid subscriptions, no ad network and no digital products; monetizing means the paid tier regardless of list size. Ghost’s entry plan no longer supports paid subscriptions at all, so charging starts one tier up.
The rubric consequence is specific. Monetization design band 1 is no commercial surface of any kind, which is exactly what a gated free tier produces: a publication with an audience and nowhere for existing demand to land. That is a scoring floor imposed by a pricing decision, not by the publisher.
Notes on each
Structural characteristics only. Prices and tier limits change and are governed by the verification date above; the shape of what each platform does with your list changes far less often.
Self-hosted sender
An application you run plus a transactional sending provider. Zero share, no bracket pricing, and you hold the database outright, so the exit cost is genuinely nothing. The cost is transferred rather than removed: updates, backups, bounce handling, suppression lists and deliverability all become your job. The publishers who do this well were going to enjoy doing it, and it is a poor recommendation for anyone else.
Ghost(Pro)
Publishing and sending as one product with no share of subscription revenue, full export, and a custom domain expected rather than upsold. Prices by total member count, which means free subscribers are billed the same as paying ones, so a large free list is a cost centre here and a rounding error on self-hosted. The wider platform trade is in Ghost vs WordPress for a one-person publication.
Kit
Built around creator commerce rather than around a publication. No share of subscription revenue, a transaction fee on product sales, own sending domain supported, full list export. Strong on the Distribution loop mechanics of capture and automation, weaker as a publishing layer, so the archive usually lives somewhere else and the two have to be kept in step.
Buttondown
Deliberately small surface. Plain, complete export, own domain, no share. The omission is commercial: there is little in the way of built-in inventory or product surface, which is a Monetization design ceiling by absence rather than by policy. For a publisher whose commercial surface lives on their own site, that omission costs nothing.
MailerLite
A general email marketing tool rather than a publishing platform. No share, low exit cost, and no publishing layer to take, which is both the reason the exit is cheap and the reason the archive has to be someone else’s problem.
beehiiv
Generous free tier, no share on the paid plans, and monetization gated behind those plans rather than owned outright. The gating is the structural issue and is covered below.
Substack
Zero bill, permanent share, and the highest exit cost on the table. Covered in full above.
The test to run before you commit
Export everything on day one, before there is anything to lose. Then check four things.
- Are the images in the file, or are they links? Open the exported HTML with your network connection off. What renders is what you actually own.
- Do the subscriber records include what you would need to rebuild the list? Addresses alone are the minimum. Names, signup dates, source, and status are what make a list a relationship rather than a spreadsheet.
- Do the paid subscriptions live in your payment processor or the platform’s? If the Stripe account is yours, the billing relationships are portable. If it is not, migrating paid members means asking them to re-enter card details, and a predictable share of them will not.
- Does the archive live on a domain you registered? This is the only one of the four that cannot be fixed later.
The fourth question decides more than the other three combined, and it is settled on the first day of the publication by whoever bought the domain, or did not.
How this is scored
Platforms are not scored in teardowns. Publishers are. But two dimensions have ceilings that a platform choice can set from the outside.
Owned hub strength band 4 requires that domain, archive and list are all owned and portable, and that analytics do not depend on the publishing platform. Band 5 adds that the archive can be rebuilt and rehosted from source without losing a single URL. An export that omits media cannot satisfy band 5, because the archive cannot be rebuilt from what you hold. A subdomain cannot satisfy it either.
Distribution loop asks whether attention becomes a relationship the publisher controls. A list that can be exported but not authenticated from your own domain is a relationship the platform is holding on your behalf. The requirements that make an owned send deliverable are on the email authentication reference page.
Sending is a commodity. Owning the list, the addresses and the data is not, and that is the whole of what separates the top of this table from the bottom.
Questions
Is a revenue share ever the right choice?
Yes, below the crossover, and the crossover arrives sooner than most plans assume. A publication with no paid subscribers pays nothing for a share and pays real money for a flat fee. The mistake is not starting on a share, it is staying on one after the arithmetic reverses, and then discovering that leaving costs more than the share does.
Does a free tier that excludes monetization matter if I am not monetizing yet?
It matters for the reason it is easy to dismiss. A publication that has an audience and no surface for demand to land on cannot discover what the audience would have paid for. The cost is not the missing revenue, it is the missing information, and it compounds for as long as the gate is in place.
Should I split publishing and sending across two platforms?
It is a legitimate architecture and the most common one among publishers who take ownership seriously: the archive on infrastructure you control, sending on a specialist. The cost is that capture, membership state and paywalling now span two systems and have to be kept in sync, which is real work and is the thing bundled platforms are selling.
What should I do on day one if I do not know which model fits?
Register a domain, point it at whatever you choose, and confirm you can export the list and the archive before you have anything worth exporting. Those three actions preserve every option. Nothing else on this page is irreversible, and the domain is the only decision that cannot be made later.
Does this ranking apply to a sponsorship-funded publication?
The share column stops mattering and the bracket column becomes the dominant cost, because a sponsorship model wants the largest possible free list and bracket pricing charges for exactly that. Read the table with the second and third columns swapped in importance.
Sources
Pricing, free-tier limits and export behaviour verified July 28, 2026. This entry is re-verified when any listed platform changes its plan structure, and the verification date above is the one that governs.