Niche Selection

#concept #positioning

Summary

Who you sell to dominates what you sell. Every source makes audience choice the highest-leverage decision — the video via "hungry crowd beats best burger", Rodenko via "80% of sales is WHO you talk to, not WHAT", and Tony most bluntly: "niche is upstream of everything" — without an ICP you can't outbound, can't productize, and can't delegate sales at all (2026-06-15-more-clients-dev-agency).

Current Understanding

The market picks the offer, you don't invent it. Don't design a novel service; find what the market is already asking for and name its bottleneck back to it (2026-07-17-design-the-perfect-offer). The burger analogy: "Is it better to have the best burger in town, or be around a bunch of hungry people?" — hungry crowd wins every time. Corollary from the fitness-coaching example: market saturation is never the real problem; the problem is the right offer in front of the wrong people.

If it won't sell, suspect the audience before the offer. "Если оффер не продаётся — почти всегда не та аудитория" (2026-06-15-selling-development-services-in-the-ai-era). Reinforced by: if even one person bought, there's no reason others won't — the problem is never the product, it's WHO and HOW. This is a useful debugging heuristic and also an unfalsifiable one; it can rationalize away a genuinely bad offer indefinitely. Hold it as a first hypothesis, not a conclusion.

Three selection rules (primary: 2026-06-15-rodenko-selling-development-expensively for 1–2, 2026-06-15-17-ways-first-client for 3):

Sell to leadership, not implementers. To an implementer your AI is a threat to their job, and you are a risk to be blocked. To leadership, you deliver a KPI. This is the vault's clearest account of why the same offer lands or dies depending on the seat it's pitched to.
Go where AI is powerless — narrow industry methodologies LLMs don't know (Rodenko: "not profitable to feed them"). Niche depth is the moat precisely because the substitute can't reach it.
Specificity beats volume — "Shopify dev for fashion brands" beats "web developer" everywhere. Against "we do everything for everyone" (also Tony's mistake #2).

Pick the segment before you build (2026-06-15-making-money-with-ai-2026): choosing a target segment is an explicit step before implementation in all three monetization pipelines — "solve a concrete pain, don't build a platform."

The buyer default at $1K+/month: business owners (2026-07-23-make-my-first-100k-in-month, dan-martell): small-business owners feel all three buyable outcomes (time, money, status — outcome-based-selling) and decide fast — no committee. A coarser cut than "sell to leadership" but the same logic: pick the seat where the pain converts to a purchase decision. Same source adds the vault's first supply-side filter, Ikigai — love it / good at it (reframed: what do people tell me I'm good at) / world needs it / will pay for it — a check on the seller, complementing this page's market-side rules; only "will they pay" overlaps the pain-discovery machinery, and the other three quadrants are motivational rather than evidential.

SOM beats TAM — the same rule from the venture side (2026-07-26-how-to-build-a-billion-dollar-company-2027, oskar-hartmann, added 2026-07-26; vocabulary defined on tam-sam-som). Investors love a big TAM, but the start must be a small market where you can take a meaningful share now: "AI agent answering calls for HVAC/plumbers/roofers" beats "AI agents for every profession"; Manifest became a unicorn on immigration law alone; Fab.com's peak was $100M on design home goods for a loyal base. His formula — big market + small winnable sub-market + MVP, not a fantasy product — is this page's specificity rule restated by a third independent tradition (VC/product world, after the US coaching and RU dev-sales schools). That makes narrow-first the vault's most independently-converged claim after the core thesis itself. One nuance he adds that the services sources don't: the narrow segment is for winning now, with the big market kept behind it — niche as a beachhead, not a destination.

Both halves of the pincer. Note that rules 2 and 3 point the same way as the productized-service commodity test: the narrower and more domain-loaded the category, the less substitutable it is, and the more pricing-from-value becomes available. Niche choice is upstream of pricing power.

Evidence

Hungry crowd vs. best burger; fitness-coaching saturation example — 2026-07-17-design-the-perfect-offer
"Listen to what the market is already asking for" — 2026-07-17-design-the-perfect-offer
"Niche is upstream of everything"; the generalist-vs-niched table; stack-marketing mistake — 2026-06-15-more-clients-dev-agency (Tony)
"80% is WHO, not WHAT"; "sell to leadership, not implementers"; "go where AI is weak" — 2026-06-15-rodenko-selling-development-expensively
"The most important thing is WHO you make the offer to"; "if one bought, others will" — 2026-06-15-konspekt-aphorisms
"Shopify dev for fashion brands beats web developer"; niche specificity beats volume — 2026-06-15-17-ways-first-client
Pick a target segment before building — 2026-06-15-making-money-with-ai-2026
"At $1K+/mo sell to business owners"; the Ikigai supply-side filter — 2026-07-23-make-my-first-100k-in-month (dan-martell)
Condensed restatement — 2026-06-15-selling-development-services-in-the-ai-era
SOM > TAM; HVAC-agent vs all-professions; Manifest immigration-law unicorn; "big market + small winnable sub-market + MVP" — 2026-07-26-how-to-build-a-billion-dollar-company-2027 (oskar-hartmann, third independent tradition)

pain-discovery — what you do once the audience is chosen
client-acquisition-channels — niche is the precondition for any channel
pricing-from-value — the "no money" objection branch is an audience-selection failure
productized-service — the commodity test is the same insight from the offer side
ai-market-shift — AI weakness is what defines the defensible niche
future-of-engineering-work — the compliant-enterprise-harness gap is a niche opportunity
overview

Contradictions / Uncertainty

"It's always the audience" is unfalsifiable as stated. Combined with "if one person bought, the product is fine", it provides a permanent excuse never to fix the offer. Neither source names the condition under which the offer is the problem.
The 80/20 WHO-vs-WHAT split is a figure of speech, not a measurement. Uncited.
Tension with ai-market-shift: "go where AI is powerless" assumes LLM capability is static. If the AI-weak niche is a moving frontier, a niche chosen on that basis has an unknown shelf life. No sales source addresses this — though 2026-07-06-sebastian-interview-ai-and-software-engineering gives the one concrete example of a durable AI-weak niche (COBOL: no training data) plus a different kind of durable niche entirely (the compliant enterprise harness, defended by regulation rather than by AI weakness — see future-of-engineering-work).

Next Questions

What is the falsification condition — how many wrong audiences before the offer is the problem?
Which industry methodologies are durably AI-weak vs. merely not-yet-covered?
Does "sell to leadership" survive in orgs where implementers hold procurement veto?

7.5 KiB Raw Blame History Unescape Escape