Files
BusinessNotes/wiki/concepts/sell-before-build.md
EugeneTes 60176d2fdc all
2026-07-30 11:15:52 +02:00

70 lines
9.5 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
# Sell Before Build
#concept #validation #strategy
## Summary
Prove demand with **paid money before building anything** — because building is the most expensive, least reversible move, and every cheaper signal lies. Founded as a page by [[2026-07-26-main-principle-of-successful-business]] ([[oskar-hartmann]]), but the rule itself is now held by **three independent traditions**: RU dev-sales ("validate market and pain before writing code"), [[dan-martell]] (the $50 paid waitlist; offer → pre-sell → build), and Hartmann (the full experiment toolkit below). That makes it one of the vault's best-converged disciplines. A 2026-07-29 short restates the rule in AI-era compression ([[2026-07-29-start-a-business-with-claude-code]]) — initially anonymous, owner-attributed same day to [[dan-martell]]: **within-author restatement, not a fourth voice** (he already counts among the three traditions); the echo-grade caution paid off, since no independence weight had been granted. What this page adds over [[productized-service]]'s validation paragraph: the **signal hierarchy**, the **toolkit**, and the **red flags**.
## Current Understanding
**The signal hierarchy** — what counts as evidence of demand, weakest to strongest:
1. **Click on a button** — weak.
2. **Email / waitlist signup** — stronger, and deceptive ("interesting!" is free).
3. **Payment** — the only real signal. *"Голос деньгами"* — voting with money.
4. **Pre-payment** — the supreme form: the customer finances your build. The humanoid-robot-data founder sold **$5M of prepaid contracts before collecting any data**; Tesla took 100K paid Model 3 preorders before building — "a $1.5T company still tests demand."
Corollary from the Zavent anti-case: if customers will *only* pay months after delivery, that's not merely weaker demand — **you have become a bank financing your clients**, which is a different business with different economics ([[unit-economics]]).
**The toolkit, ordered by cost:**
| Experiment | Mechanics | Cost | Caveat |
|---|---|---|---|
| AI-mockup + priced offer | Generate the most beautiful image of the thing; send to the target community with a real price and "who's with me?" | ~Nothing | Hartmann's coworking test: ~5,000 reached, 0 buyers, project killed pre-build |
| Landing → **payment screen** | A plain landing page doesn't qualify; the test ends at a payment attempt | Small | One word in the offer can move conversion 35× |
| Fake payment screen | Full-looking app; at payment: "Error, payment failed"; keep intent data | <$1,000/app | **Deceptive** — see Contradictions |
| Fake-door buttons | Ship all features as non-working buttons; build in click-count order (Samwer's eBay clone) | Small | Click is the *weakest* signal — use for prioritization, not demand proof |
| Wizard-of-Oz manual MVP | Deliver the value **by hand for the first 10 clients**, then automate | Labor only | The honest workhorse for services; Builder.ai is both its exemplar and its cautionary tale (see source page) |
| Visible pre-order / paid waitlist | Charge (even partially) before building — Martell's $50 top-of-waitlist slot | ~Nothing | The honest high-signal variant |
**Why cheap and frequent beats right:** intuition always deceives (the app studios found "quit casino" dead and "quit porn" live; Samwer's click order was "never what you'd guess"), so the discipline is many small falsifications, not one confident bet. The 12-week portfolio contrast: team A ran 10 tests and had constant customer signal; team B was still building with zero — same money, "the only difference is the willingness to launch ugly and look stupid." Failing at $1K is safe; failing after a 6-month build is fatal. This is [[sales-discipline]]'s reps-not-views logic applied to *what to build* rather than *how to market*.
**The red flag inverted:** "we've been working on it since 2016, no revenue" reads as dedication and is disqualifying — duration without revenue is absence of evidence. Cross-tradition convergence: Rodenko's *"a year of repackaging is for people afraid to pick up the phone"* is the same claim from the services side ([[sales-discipline]]).
**The AI-era compressed variant** ([[2026-07-29-start-a-business-with-claude-code]], added 2026-07-29): stand up the *appearance* of a business with [[claude-code]] (landing page, invented company name, waitlist), close paying customers by cold outbound, and only then have the AI build the product. Two points of contact with this page's machinery: (1) its internal waitlist is **unpaid** and so fails the vault's own bar — but the playbook's actual validation event is the **closed sale**, which sits above mere payment in the hierarchy, so it passes the money bar where it counts; (2) the invented-company front belongs with the toolkit's deception-based rows (a fake *business* rather than a fake payment screen) — same signal logic, same unmodeled reputational/legal exposure. Author: [[dan-martell]] (attributed 2026-07-29) — which makes the unpaid waitlist a **within-author drift** from his own $50-paid-slot rule, logged on his page. 42 seconds, no case.
**Boundary with the vault's other machinery:** [[pain-discovery]] finds *which* pain might pay (talk, watch spend); this page is the *experimental confirmation* step (make them pay, small). The acute-pain test bridges the two: if people pay even when the product is bad (hospitals), the pain is acute enough to sell a concept. Then [[productized-service]] packages what the test proved, and customer cash funds the build (Martell's format-triage logic — same sequence).
## Evidence
- Signal hierarchy, all toolkit rows, coworking/Zavent anti-cases, Tesla/robot-data/Samwer cases, 12-week contrast, red flag, checklist — [[2026-07-26-main-principle-of-successful-business]] ([[oskar-hartmann]])
- $50 paid-waitlist pre-sell; offer → pre-sell → build sequence; "customer cash funds the product" — [[2026-07-23-make-my-first-100k-in-month]] ([[dan-martell]])
- "Validate market and pain before writing code"; "most AI projects die from idea → code → launch" — [[2026-06-15-making-money-with-ai-2026]] (RU tradition)
- Convergent red-flag from the services side: "год перепаковки — для тех, кто боится взять трубку" — [[2026-06-15-rodenko-selling-development-expensively]] via [[sales-discipline]]
- Within-author restatement ([[dan-martell]], slogan-grade): sell via cold outbound before any product exists; "then ask Claude to build it" — [[2026-07-29-start-a-business-with-claude-code]]
## Related Pages
- [[productized-service]] — what gets packaged once the test pays; holds the when-to-productize dispute
- [[pain-discovery]] — upstream: finding the candidate pain; the acute-pain test lives on both pages
- [[unit-economics]] — pre-pay vs post-pay decides whose balance sheet carries the build
- [[sales-discipline]] — the same anti-perfectionism discipline, marketing-side
- [[cloning-over-originality]] — the toolkit itself is cloned (Samwer → Hartmann; 10 LLM-data companies → robot-data founder)
- [[niche-selection]] / [[tam-sam-som]] — "the smallest group with the most acute pain, by name" is a SOM statement
- [[oskar-hartmann]], [[overview]]
## Contradictions / Uncertainty
- `Status: tentative` on the specific mechanics (single-source, war-story figures); the *rule* itself (money before build) is three-tradition and as solid as anything in the vault.
- **Two toolkit rows are deception-based** (fake payment screen, fake-door buttons): they collect purchase intent under false pretenses, carry consumer-protection exposure in some jurisdictions, and their reputational cost if discovered is unmodeled by the source. The honest variants (visible pre-order, paid waitlist, manual MVP) deliver most of the same signal; the vault records the deceptive mechanics without endorsing them. A third deception-shaped variant arrived 2026-07-29: the invented-company front ([[2026-07-29-start-a-business-with-claude-code]]) — same caveat applies.
- **Tension with the articulation thread:** "don't dictate what the customer should want — build what's demanded" sits against [[outcome-based-selling]]/[[pain-discovery]]'s claim that naming a pain *better than the buyer can* is the value-add. Likely scoped — product demand testing vs service framing — but no source draws the line.
- The traffic-arbitrage trainer recommendation doesn't transfer cleanly to services (no cheap payment-screen A/B); the transferable core is "smallest paid test."
- Survivorship in the positive cases: Tesla and the $5M founder are selected wins; the fake-payment studios' failure rate across their 20-app batches is exactly the data not shared.
## Next Questions
- ~~What is the smallest *paid* test for a dev-services offer — a paid audit? a deposit-backed discovery sprint?~~ **Proposed answer 2026-07-26** ([[2026-07-26-eugene-90-day-plan]]): a **fixed-price, fixed-scope paid diagnostic** scoped to *one* line/device/process. It clears four constraints at once — money before build, small enough to skip procurement, produces the missing case study, and doubles as an [[offer-ladder]] entry rung. Vault inference, **untested**; no source states it. Its live falsifier: if free feasibility assessments are the industrial norm, the paid bar can't sit here and must move downstream.
- Where is the legal line for fake-door testing in the owner's jurisdiction(s)?
- Does pre-payment willingness vary by market culture (US vs EU vs CIS B2B), and does the Zavent post-pay trap generalize to enterprise services where net-60/90 is standard practice?