customer-blueprint — the customer stone

Every business claims to know its customer; almost none can quote one. The usual persona deck — “Marketing Mary, 34, lives in a suburb, likes yoga” — is demographics cosplaying as understanding, and nothing downstream can use it. What messaging, pricing, and go-to-market actually need is psychology and language: what the customer is afraid of, what they’re reaching for, what finally makes them buy, what makes them balk, and the exact words they use for all of it. This skill writes that down, interviews the founder against real evidence, refuses to launder guesses into quotes, and ships the result as a Business OS module: canonical CUSTOMER.md, machine contract, core tables, MCP server, standalone site.

This is stone 2 of 8. Read ../PROTOCOL.md before composing — the module contract, workspace layout, and definition of done live there and are not repeated here.

When to use it

The user says any of: “customer blueprint”, “customer personas”, “ideal customer profile”, “who is our customer”, “target audience doc”, “/customer-blueprint” — or invokes messaging, business-model, or go-to-market in a workspace with no customer-blueprint/ module (offer this first; don’t force it).

Upstream: locate vision/VISION.md in the workspace and extract from it — especially §05 (what we refuse) and §06 (positioning) — instead of re-asking. If it’s missing, flag loudly per protocol: “composing the customer blueprint without a locked vision — segment refusals are provisional.”

The rubric — sections of CUSTOMER.md

Compose in order; each section feeds the next. Propose → confirm → lock, one section per exchange (the interview mechanics in ../../potatuhs-design/SKILL.md §Procedure apply verbatim).

00 · Cover — business name, one-line statement of who this business is for, persona count, version, date.

01 · Segments served & declined — the market cut into named segments; which ones this business serves and, just as deliberately, which it declines. Every declined segment names the cost of serving them and cites the vision’s refusals (§05) where they align. A blueprint that declines nobody is a mailing list, not a strategy.

02–0N · One section per persona — one section per persona, each built on the same seven dimensions: fears (what keeps them up at night), motivations (what actually drives the purchase), aspirations (who they’re trying to become), frustrations (what the status quo does to them daily), buying triggers (the events that turn latent pain into a purchase), objections (what stops them at the point of decision, with the honest answer to each), and vocabulary (the words they use for the problem — not the founder’s words). Name each persona after the role or situation, not an invented human. Two or three deep personas beat six shallow ones; push back on persona sprawl.

0N+1 · Voice of customer — the consolidated vocabulary bank: real phrases customers have said, each attributed to a persona and a source (interview, support ticket, review, sales call). This is the section messaging mines directly — headlines get written from these rows. Thin evidence here gets a loud placeholder, never invented dialogue.

0N+2 · Buying journey, triggers & objections — how a customer moves from unaware to bought: the stages, what triggers movement between them, where each persona typically enters, and where deals die. The objections from the persona sections roll up here with the business’s committed answers — go-to-market sequences its plays against this map.

0N+3 · Research method & evidence log — how customer research happens ongoing: interview cadence (e.g. n per month, who conducts them), the question spine, and where evidence lands — every interview gets a row in the interviews table, and personas get revised from evidence, never from vibes. A blueprint without a research method is a snapshot pretending to be a system.

The module

Per ../PROTOCOL.md (order 2, port 4002, upstream reads: vision):

• CUSTOMER.md — the canonical document, sections above.
• customer-blueprint.json — contract: { name, forStatement, segments{ served[{name, why}], declined[{name, cost, visionRefusal}] }, personas[{id, name, fears[], motivations[], aspirations[], frustrations[], triggers[], objections[{objection, answer}], vocabulary[]}], voiceOfCustomer[{phrase, personaId, source}], journey{ stages[], triggers[], dropoffs[] }, research{ cadence, method, evidenceHome } }.
• Core tables — personas (id, name, segment, status, revised_on, evidence_count): the living register the document’s persona sections project from, so a persona’s evidence base is queryable, not asserted. interviews (persona_id, date, source, quote, insight, links): the evidence log — every conversation lands here first, and section 0N+3’s cadence is measured against its row count.
• MCP — the four standard tools (customer-blueprint_get_document, customer-blueprint_get_contract, customer-blueprint_query, customer-blueprint_record — the write path for new interviews).
• Site — single-page-scroll document site, one section per rubric entry, per the architecture contract in ../../potatuhs-design/SITE.md (sidebar nav, scroll-sync, print-to-PDF). If a brand stone exists in the workspace, render with its tokens; otherwise ship the neutral default theme and note it in status.
• CLAUDE.md — states that messaging, go-to-market, and business-model read this document; a change to any persona, the voice-of- customer bank, or the journey map requires checking all three for stale citations before the session ends.

Hard rules

• Personas are evidence-backed. Every persona claim traces to something a real customer said or did. Where evidence doesn’t exist yet, the persona is a hypothesis and the document says so — hypotheses are honest; disguised guesses are not.
• Real quotes only. An invented customer quote is a fabrication, and it’s the worst kind — it poisons messaging downstream with words no customer ever used. No interviews yet means an empty voice-of-customer section with a loud placeholder and a research cadence to fill it.
• The vocabulary section records what customers actually said — their phrasing, their metaphors, their misuse of the founder’s terminology. Correcting their words into industry language destroys the section’s value.
• Declined segments must decline something real — a segment with actual revenue attached and a stated cost of serving it, citing the vision’s refusals where they overlap. Push once; if the founder holds, take their answer.
• No demographics-as-insight. Age, city, and job title may appear as context; they never substitute for the seven dimensions. A persona defined only by demographics is a placeholder, marked as one.
• Placeholders over fabrications, loudly, per protocol.