BusinessNotes/wiki/sources/2026-07-26-main-principle-of-successful-business.md

Главный Принцип УСПЕШНОГО БИЗНЕСА (The Main Principle of a Successful Business)

#source #validation #strategy

Source Metadata

• Raw path: raw/sources/Главный Принцип УСПЕШНОГО БИЗНЕСА.md
• Source type: Conclusions note from a Russian-language YouTube video (17:52), https://www.youtube.com/watch?v=pHoWRl7MVwE
• Speaker: oskar-hartmann (named in the raw note) — his second vault source
• Date: undated (2026 context — AI mockups, humanoid-robot data market)
• Ingested: 2026-07-26

Core Claims

1. Sell first, then build. Building the product is the entrepreneur's most expensive and least reversible decision; before committing, prove with cheap experiments that the pain is acute enough that customers pay money up front. Waitlists, likes, "interesting" — none of it counts (sell-before-build).
2. The signal hierarchy: button click (weak) < email/waitlist signup (stronger but deceptive) < payment — the only real signal. "Голос деньгами" — voting with money.
3. The cheap-experiment toolkit:
   • AI-mockup blast (his own anti-case): a beautiful AI-generated image of a coworking space for business owners, offered to ~5,000 people via 7–8 business clubs with named price and "20 seats, who's with me?" — 0 real buyers despite his 100% conviction. Project killed for pennies instead of millions.
   • Fake payment screen: top utility-app studios launch ~20 apps at once as full-looking shells; at the payment screen the user sees "Error, payment failed" — the studio keeps real willingness-to-pay data at < $1,000 per experiment. Findings: "quit casino" failed, "quit porn" worked, weight-loss always works. The studios forbid themselves from inventing — they let money vote.
   • Fake-door buttons (Oliver Samwer's eBay clone for Germany, ~20 years ago): shipped all 100 buttons in the UI non-functional, watched click logs, built in click-count order. "Intuition deceives — always."
   • Wizard-of-Oz manual MVP: when the product can't be cheaply mocked, deliver the value by hand for the first 10 clients, then automate. (His example: Build.ai [sic] — valued at $2B while "the AI" was 500 Indian specialists; "great way to start — at $2B you should probably have built the product.")
   • Pre-orders at any scale: Tesla collected 100K paid Model 3 preorders, then built; same for Cybertruck — "a $1.5T company still tests demand; the principle doesn't depend on size."
4. Pre-payment is the supreme demand signal. The MIT-graduate case: sold $5M in prepaid contracts for humanoid-robot training data ("fold shorts + sensors + video") before collecting any data — strategy copied from 10 companies that grew the same way on LLM data. Positive unit economics from day one. Anti-case: his own Zavent (meeting-room booking) — customers would only pay 3 months post-delivery, turning the company into "a bank financing its clients"; when prepayment was required, conversion collapsed. If customers will pay up front, demand is real; if only post-pay, you're in a different business.
5. The acute-pain test: "If the business only works when everything is perfect — it's a bad business. A good business is when everything is bad and people still come and pay." Exemplar: hospitals — 1.5-hour waits, no good reviews, you pay anyway and say thank you. Acute pain means people pay for the concept, not for polish (pain-discovery).
6. The recommended trainer business is traffic arbitrage: fastest feedback loop — buy traffic → landing → payment screen; a single word in the offer can move conversion 3–5×. Rule: a plain landing page doesn't qualify — there must be a payment screen.
7. The mindset differential: two portfolio AI companies, 12 weeks, same money — team A ran 10 tests, updated the product repeatedly, constant customer signals; team B, still building, zero signals. Only difference: willingness to launch ugly and look stupid. "Success is usually not where you thought — and only frequent cheap tests find it."
8. Red flag: duration without revenue. "We've been working on this since 2016, no revenue yet" is pitched as dedication and is "the fattest minus" — no revenue over a long project = no evidence. "I'd rather talk to a team that started a week ago."
9. CAC-burning as a moat, restated: once PMF is proven and unit economics are positive, spend heavily per customer ($80–100 CAC) so no newcomer can afford to enter — consistent with his first source's ⅓-of-LTV entry-barrier logic (unit-economics, sales-channel-as-moat).
10. Через-principles: experiment while it's cheap ($1K mistake safe; 6-month build fatal); intuition always deceives; endure the pain of reality early; don't dictate what the customer should want — build what is already demanded.

Key Evidence / Details

• The self-test checklist (raw note): smallest experiment that would falsify the hypothesis · what makes a customer pay now, pre-build · landing → payment screen (not signup) · can I collect (partial) prepayment · who — by name — is the smallest group with the most acute pain · will I walk to 10 concrete potential buyers and ask what they'd pay for right now · can I deliver the value by hand to the first 10 · if >3 months with no revenue, what actually stops an ugly launch tomorrow · does my business survive when everything goes wrong?
• Named cases: coworking test (own, failed cheaply), utility-app studios, Samwer/eBay-Germany, Tesla Model 3/Cybertruck, humanoid-data founder ($5M prepaid), Zavent (own, post-pay trap), Build.ai [sic].

Connections

• sell-before-build — new page; the whole discipline (signal hierarchy + toolkit) lives there
• productized-service — validate-before-build gains a third tradition; the $50-paid-waitlist and RU validate-first rules now have a full toolkit behind them
• pain-discovery — the acute-pain test ("still pay when everything is bad") is a new, revealed-preference diagnostic
• cloning-over-originality — first independent second voice: Samwer's clone, app studios that forbid inventing, the data founder copying 10 LLM-data playbooks
• sales-discipline — "since 2016, no revenue" red flag converges with Rodenko's "a year of repackaging is for people afraid to pick up the phone"; welcome bad news early
• unit-economics — Zavent: post-payment = financing your customers (a forgotten cost line made vivid); CAC-burn barrier restated
• sales-channel-as-moat — within-author consistency on the CAC entry barrier
• oskar-hartmann — second source; framework stability
• eugene — the checklist is directly runnable against his CV/embedded offer

Open Questions

• Ethics/legality of the deception-based tests are never addressed. Fake payment screens ("error, payment failed") and fake-door buttons collect purchase intent under false pretenses — consumer-protection exposure varies by jurisdiction, and the reputational cost of a discovered fake is unmodeled. The vault records the mechanics without endorsing them; the honest variants (visible pre-order, paid waitlist, manual MVP) carry most of the same signal.
• "Build.ai" is almost certainly Builder.ai (general knowledge, not from the source): the widely reported case matching the description — and that company collapsed into insolvency in 2025 amid allegations its AI was largely manual work and its revenue overstated. The source cites it as a quasi-positive "great way to start," which undercuts the example: the same case is also the cautionary tale for never graduating from Wizard-of-Oz (and for misrepresenting it to investors — the part Hartmann's framing skips).
• All figures are war-story grade: 5,000 recipients / <5 replies, <$1K per app test, 100K preorders, $5M prepaid, $80–100 CAC, 3–5× one-word conversion swings — attributable, unverified.
• The traffic-arbitrage recommendation sits oddly with the vault's services domain — a services operator can't A/B a payment screen as cheaply; the transferable core is "smallest paid test," not the mechanism.
• Tension worth watching: "don't dictate what the customer should want" vs the vault's outcome-based-selling/technical-founder-trap thread, where articulating an outcome the buyer couldn't name is the value-add. Likely scoped (product demand vs service framing), unstated.

Change Impact on Wiki

• Created sell-before-build (concept).
• Updated productized-service (pre-sell now three traditions), pain-discovery (acute-pain test), cloning-over-originality (first independent second voice; status upgraded), sales-discipline (duration-without-revenue red flag convergence), unit-economics (Zavent post-pay trap; CAC-burn consistency), sales-channel-as-moat (consistency note), oskar-hartmann (second source), eugene (runnable checklist), overview, index, log.