Should a Marketplace Build for One Supplier First? How Substack Started With Bill Bishop
Short answer: Build around one supplier first when that person represents a reachable segment, already has buyer trust and forces the complete transaction to work. Serve them deeply enough to process real sales, but label every special request and refuse to call it a platform until two more independent suppliers succeed through substantially the same process. One impressive launch proves possibility, not repeatability.
A marketplace appears to need both sides at once. That belief pushes founders to recruit hundreds of suppliers before knowing whether one can complete the promised transaction.
Starting with one supplier avoids empty breadth. It creates a different risk: the product becomes private software for an exceptional person whose audience, reputation and needs do not represent the next customer.
Substack began with a known professional writer
Substack's founders initially focused on professional writers rather than everyone who might send an email. A company-published account of its early community says they looked for writers who could follow the reader-funded model demonstrated by Stratechery's Ben Thompson.
Co-founder Hamish McKenzie already knew Bill Bishop, author of the China newsletter Sinocism. Bishop had written it free for five years and had told readers that he planned to introduce a paywall. McKenzie proposed that he become Substack's first publisher.
McKenzie and Chris Best flew to Washington, DC, and built the first version around Bishop's needs. Substack's four-year retrospective with Bishop says he enabled paid subscriptions on 15 October 2017 and brought in six figures of revenue by the end of that day. The source does not disclose the exact amount, subscriber count, price mix or refunds.
The launch proved that a trusted writer with an established audience could use an integrated publishing and payments service to sell subscriptions. It did not yet prove that an unknown writer could build an audience, or that Substack could acquire writers economically.
My view is that a marketplace should earn the right to recruit supply in batches. One supplier should expose the whole system. Three similar suppliers should show whether the system travels.
Use the One-Supplier Proof Sequence
The One-Supplier Proof Sequence separates a useful design partner from misleading special treatment.
| Proof | What one supplier must reveal | Passing evidence | |---|---|---| | Representation | The supplier resembles a segment you can find again | Clear shared job and reach route | | Transaction | Buyer can discover, pay and receive value | Real completed purchases | | Economics | Revenue covers direct delivery and support | Positive or deliberately bounded contribution | | Transfer | A second and third supplier can repeat it | Same core workflow with less founder work |
Do not start transfer work before the first transaction completes. Do not claim transfer until unrelated suppliers use the core process.
Choose a supplier with evidence on both sides
The best first supplier has an urgent problem and credible access to buyers. Existing audience is one form of access, but not the only one. A respected specialist may have a client list, professional community, retail footfall or recurring referrals.
Avoid selecting only for fame. A famous supplier may convert because of reputation unavailable to the wider segment. Write down which assets belong to the supplier and which the marketplace provides.
Bishop brought years of free publishing and reader trust. Substack contributed the paid publishing workflow. Keeping those contributions separate prevents the platform from crediting itself with all demand.
Complete the transaction manually where necessary
Map every step from supplier onboarding through customer payment, delivery, cancellation and payout. Early founders can handle some steps manually to learn, provided customers receive what was promised and regulated activities are handled lawfully.
Record each intervention. Did the founder import an audience, format the offer, answer support or reconcile payouts? Work that occurs once may be onboarding. Work that recurs per buyer can destroy the model at scale.
Payments, tax, consumer subscriptions, data protection and supplier classification vary by country. Use regulated payment providers where required and obtain qualified local legal, payments and tax advice.
Separate configuration from custom development
A first supplier will expose missing features. Classify each request as core, configurable, segment-specific or unique.
Core work is necessary for every transaction, such as payment confirmation. Configurable work lets suppliers choose a price or branding within a common system. Segment-specific work may deserve a later product. Unique work stays manual, paid separately or declined.
If the first supplier determines the database, interface and commercial terms through unique requests, the marketplace is becoming an agency. That can be a valid business, but it is a different one.
Worked example: PolicyLetter
PolicyLetter helps independent analysts sell paid briefings. These figures are illustrative.
Its first supplier, Maya, writes a free UK planning-policy email to 8,000 readers. She charges £120 a year for a weekly paid briefing. In the first month, 160 readers pay, producing 160 x £120 = £19,200 in annual subscription sales.
PolicyLetter charges 10%, so gross platform revenue is £19,200 x 10% = £1,920. Payment fees are 2.4% plus 20p per transaction: £19,200 x 2.4% = £460.80, plus 160 x £0.20 = £32, for £492.80. PolicyLetter has agreed that those fees come from its share. Founder onboarding and support take 38 hours at £35, costing £1,330.
First-launch contribution is £1,920 - £492.80 - £1,330 = £97.20 before central overhead.
| Work item | Maya | Second analyst | Third analyst | |---|---:|---:|---:| | Founder onboarding hours | 38 | 19 | 11 | | Paid readers in first month | 160 | 42 | 35 | | Refund rate | 3.1% | 2.4% | 2.9% | | Platform contribution | £97 | £184 | £201 |
Maya's launch proves payment and delivery but barely contributes because founder work is high. The next two analysts have smaller audiences, yet require less setup and leave more contribution. Their combined 77 buyers show that the workflow is not unique to Maya.
PolicyLetter sets its batch gate: onboard five planning and local-government analysts with fewer than 20 founder hours each, at least 25 paid readers and positive contribution after refunds. It does not recruit lifestyle writers because their audience, cadence and support needs may form a different segment.
Measure supplier success and marketplace success separately
Supplier sales matter, but the platform also needs repeatable acquisition and delivery. Track supplier invitations, qualified conversations, launches, buyer conversion, refunds, payout accuracy, founder hours and platform contribution.
A supplier can earn substantial revenue while the platform loses money supporting them. Conversely, a profitable platform fee can coexist with a poor supplier outcome if few readers convert. Both sides must receive enough value to remain.
Use cohorts. The first supplier will almost always receive more founder attention than later ones. The key improvement is not a larger launch every time. It is a reliable launch with fewer exceptional interventions.
Expand by similarity before breadth
Recruit the next suppliers because they share the first supplier's job, not because their names look impressive. Keep the product, commercial terms and onboarding substantially stable for three launches.
When the same obstacle appears twice, improve the standard system. When only one supplier requests something, wait. This discipline prevents an early roadmap from becoming a collection of personal favours.
The guide to [finding your first profitable customer segment](/market-research/first-profitable-customer-segment/) can help define how narrow that first group should be.
Run the sequence in eight weeks
In week one, identify ten suppliers who already serve a similar audience and choose one with a live paid intention. Agree the outcome, price, responsibilities and evidence you may analyse.
During weeks two and three, make the complete transaction work. Record every founder intervention. Launch to real buyers in week four and follow refunds, delivery and support.
Use weeks five to eight to onboard two comparable suppliers without rebuilding around them. At the end, choose among a repeatable segment, a bespoke service or a failed market assumption. Do not add hundreds of profiles to disguise an unproven transaction.
Related guides
Frequently asked questions
Does the first supplier need an existing audience?
They need credible access to likely buyers, which may be an audience, customer list, physical community or professional network. An established audience speeds the transaction test but can exaggerate the platform's acquisition value. Record how buyers arrived and what trust existed before the platform. If you eventually intend to help suppliers build demand from zero, test that as a separate job. One supplier with loyal followers cannot prove that promise. Choose the first case that isolates the uncertainty you most need to resolve.
Should I offer the first supplier free use?
Usually charge normal commercial terms or a clearly bounded founding arrangement. Free use can be reasonable in exchange for substantial design access, but it does not test willingness to pay. State the duration, support and later fee before work begins. Avoid promising permanent special treatment that later suppliers cannot receive. Even when the platform fee is waived, calculate the fee that would have applied and the full founder cost. You need to know whether a repeat customer could be served economically, not merely whether the first supplier appreciates help.
How much custom work is acceptable?
Enough to discover the common transaction, but not so much that the result cannot travel. Tag every request as core, configurable, segment-specific or unique, and record hours. A working guide is to require founder intervention to fall meaningfully by the second and third supplier. If each launch demands a fresh product, you are operating a service business. That may be profitable, but price and staff it honestly. Do not call bespoke implementation a scalable marketplace simply because buyers and sellers use software.
What if the first supplier fails to sell?
Separate audience, offer, price and transaction failure carefully. Check whether qualified buyers saw the offer, understood it, attempted payment and received the promised value. Interview non-buyers about actual alternatives, not general interest. If the supplier lacked buyer access, choose a stronger representative before condemning the model. If trusted buyers saw a clear offer and would not pay, change one assumption and retest within a fixed spending limit. Do not recruit many weak suppliers to average away a failed first transaction.
When can I call the model repeatable?
After at least two additional, independent suppliers complete substantially the same transaction with less founder intervention and demonstrably acceptable economics. Three cases are still early evidence, not certainty, but they expose whether the first was exceptional. Define the shared segment, price route, onboarding steps and success threshold before recruiting the later suppliers. Repeatability means the system survives a genuinely new relationship. It does not mean every supplier earns the same revenue, because audience size and offer quality will vary materially.
How do I avoid becoming dependent on one star supplier?
Cap special development, document the common process and recruit similar suppliers before the first represents most platform activity for long. Model the effect of that supplier leaving. Protect data portability, payout accuracy and contractual clarity for both sides. Do not obstruct departure to manufacture dependence. A star can provide credibility and learning, but concentration weakens the marketplace's bargaining position and revenue stability. Track the share of transactions, revenue and support attached to the largest supplier and set an internal concentration alarm.
Comments (0)