Viability blueprint: ten verticals, interviewed in order, ending in a blunt verdict

---
name: viability-blueprint
description: Runs a structured pre-launch viability interview across ten go-to-market verticals in a fixed order, from the offer and the audience of one through use cases, narrative angles, competitors, content, keywords and growth loops to a blind-spot sweep, then produces a marketing readiness blueprint ending in three strengths, three bluntly stated gaps, five prioritised actions and one viability verdict. It enforces an interview discipline of one to three questions at a time, a sentence template for use cases, and a five-stage specification for any claimed growth loop. This skill should be used when assessing a product idea, a pre-launch plan or a post-pivot go-to-market before resources are committed, and when the requested output is a blueprint and a verdict rather than a critique of an existing document.
---

# Viability blueprint

## The claim this skill is built on

A pre-launch assessment is an interview problem before it is a framework problem.

Frameworks for this are not scarce. What goes wrong is the conversation. The questions get delivered as a form, the founder answers the whole form in one enthusiastic paragraph, the vague answers are accepted because probing them feels rude, and wherever an answer is missing the assessor quietly supplies a plausible one from their own head. What comes out is a fluent document about a product the founder does not entirely recognise, in which every gap has been smoothed into a gentle suggestion.

So the parts that matter here are the parts a framework does not specify: the order the verticals run in, how many questions you are allowed to ask at once, what a vague answer looks like and what you say to it, and the fixed shape of the report at the end. The ten verticals are ordinary. The discipline around them is the skill.

One boundary, stated up front. This produces a specification, not a measurement. Where a vertical asks for a growth loop, the deliverable is the loop written out with all five of its stages named. Computing its branching factor and cycle time requires a running product and belongs to a different job.

## Phase 1: understand before advising

Do not open with the framework. Ask the founder to describe the idea in their own words, with no structure imposed, and let them finish.

Then ask follow-ups until you can state four things without guessing:

1. What it does, mechanically, in one sentence a stranger could repeat.
2. Who it is for, as a role in a described kind of organisation rather than a market.
3. What problem it solves, phrased as something that happens today and costs something.
4. What stage it is at.

Stage matters because it decides which verticals can be answered with evidence and which are hypotheses. Use these five labels and record which one applies:

| Stage | What can be evidence | What is still a hypothesis |
| --- | --- | --- |
| Idea only | Nothing about demand | All ten verticals |
| Prototype, no users | Feasibility | Offer, audience, use cases, everything downstream |
| Private beta, under ten users | Use cases and activation | Pricing, channels, loops, retention |
| Launched, under fifty paying | Offer, pricing, first segment | Loops, expansion, keyword position |
| Launched, over fifty paying | Most of it | Loops, unless already instrumented |

**Then play it back in two to three sentences and get explicit confirmation.** Not a summary of the conversation, a statement of the product: what it does, for whom, against what alternative, at what stage. Ask directly whether that is right, and correct it until they say yes.

This step looks like a courtesy and it is not. Everything in phase two is checked against this statement. If it is wrong by ten per cent, the whole assessment is wrong by ten per cent in a way nobody will notice until the report lands and the founder says that is not really what we do.

## Phase 2: the ten verticals, in order

Work through all ten. For each one: give a single line on why the vertical matters, ask its questions one to three at a time, probe every vague answer before moving on, then close the vertical with a two to four sentence synthesis naming what is strong and what is missing. Then move to the next.

Why this order, since a different order is always tempting:

| # | Vertical | Why it sits here |
| --- | --- | --- |
| 1 | The offer | The only vertical answerable with no research. It becomes the object everything else is checked against |
| 2 | Audience of one | A single specific person is easier to reason about than a category, and the categories are generalised outward from them |
| 3 | Expanded personas | Generalisation only works once the specific case exists, otherwise you get demographic buckets |
| 4 | Use-case library | The concrete instances of the offer meeting the persona. Everything downstream is made out of these |
| 5 | Narrative angles | A story assembled from invented cases is fiction. Cases first, story second |
| 6 | Competitors and positioning | A comparison is meaningless until you know which jobs you are comparing on, which is vertical four |
| 7 | Article slate | Inherits the positioning. Written before it, it argues for a position you have not chosen |
| 8 | Organic keywords | Inherits the positioning and the use cases, which are where the pain language comes from |
| 9 | Growth loops | A loop compounds an existing channel. Specified before the channel exists, it is a wish |
| 10 | Blind spots | The sweep for everything the first nine did not cover, which requires the first nine to be done |

### 1. The offer

Five items: the offer in one sentence including the promised outcome, why it beats the alternatives, the pricing model, the risk-removal elements such as a pilot, a guarantee, month-to-month terms or migration help, and the contrast against the baseline, meaning what the buyer does today and what that currently costs them.

Vague signal: an offer sentence containing no outcome, only a description of the software.

### 2. Audience of one

One specific person. A role, in a described organisation, with a name you invent if it helps. Then: the pain they wake up with, the transformation they want, the beliefs that block them from buying, and the metric they are personally judged on.

Probe for the metric with "which number appears in their performance review", because that is the number that gets your invoice approved.

### 3. Expanded personas

Two to four, no more. Each with goals, actual workflow, pains, emotional triggers and budget authority. Budget authority is the one that earns its place: it routinely reveals that the buyer is three people, an evaluator, an approver and a blocker, and that the blocker has never been considered.

Vague signal: four personas with identical pains. That is one persona wearing four job titles.

### 4. The use-case library

Five to ten repeatable cases, each in this exact form:

**If you are X and struggle with Y, this solves Z.**

- X is an observable role plus situation. Not an adjective.
- Y is a behaviour or a cost that recurs on a known cadence.
- Z is a specific change with a result somebody could check.

| Fails the template | Passes the template |
| --- | --- |
| If you are a growing team and struggle with efficiency, this solves your workflow | If you are a support lead with two people covering four products, and you rewrite the same answer more than twice a week, this turns the resolved thread into a draft article in one click |
| If you are a modern retailer and struggle with data, this solves reporting | If you run a shop with more than two hundred lines and one merchandiser, and you rebuild the same stock spreadsheet every Monday, this produces it from the till exports overnight |

Then mark each case as **instant return** or **delayed return**. Instant means the buyer sees the value inside the first session. Instant cases become the demo and the onboarding path. Delayed cases become the case studies.

If no case is instant, stop and record it. An activation problem discovered at this point is the single most valuable output of the whole exercise, and it is invisible if you only read the list as a set of individually plausible benefits.

### 5. Narrative angles

Five: the origin story, the contrarian take, the how-I-solved-this, the before and after, and the data-backed prediction. Each must contain a claim a reasonable person could disagree with. An angle nobody can argue with is a description.

### 6. Competitors and positioning

Strengths and weaknesses from the buyer's view, not the product team's. The product team's list of competitor weaknesses is usually a list of things the buyer does not care about.

Include the alternatives that are not products: doing nothing, a spreadsheet plus a person, a contractor, an adjacent tool used sideways. In most young categories those outrank every named competitor combined.

Then the angle of attack, and a tagline that is either analogy-based, of the form "the X for Y", or category-breaking, which asserts the existing category is the wrong shelf. Note which one you chose, because a category-breaking tagline commits you to explaining a thing the buyer has no budget line for.

### 7. Opinion-led article slate

Three to five pieces where the product fits the story naturally rather than being bolted on. Each needs: the opinion in one line, the keyword intent behind it, meaning informational, commercial investigation, or transactional, and a distribution plan naming where the first hundred readers come from. "We will post it" is not a distribution plan and should be recorded as a gap.

### 8. Organic keywords

Three buckets: pain-based phrasings people type before they know a solution exists, solution-based phrasings once they do, and long-tail questions lifted from forums and community threads in the exact words used there.

Rank them by rankability using a heuristic that needs no tool. Look at the first page of results and count how many belong to organisations whose entire business is this topic. Eight or more means a head term you will not reach for years. Three or fewer, with forum threads and question pages ranking, means it is reachable by a good article. Say which bucket each term is in. A slate of head terms only is a gap, not a strategy.

### 9. Growth loops

Write each candidate loop with all five stages named:

1. **Trigger.** What starts a cycle, and how often.
2. **Action.** What the user does.
3. **Reward.** What they get, and whether it arrives in the same session.
4. **Shared output.** The artefact a non-user sees, named specifically, with the brand visible on it.
5. **New users.** The route from seeing that artefact to signing up, named as a path a person walks.

Stages four and five are the ones that get skipped, and a loop missing either of them is not a loop. Ask two supporting questions: what visible artefact carries the brand into places you do not control, and what collaborative workflow makes sharing the path of least resistance rather than an act of generosity.

If a stage cannot be named, record the loop as "not a loop yet" and move on. Do not repair it during the interview.

### 10. Blind spots

Seven prompts, asked one at a time: channels beyond your default; partnerships and integrations that borrow an existing audience; the easiest first segment, meaning the one that would say yes fastest even if it is small; the pre-launch mechanism, such as a waitlist, design partners or a founding cohort; the activation steps between signup and first value, counted; the retention loop that brings someone back in week three; and the pricing experiments to run before locking a number.

## Interview discipline

These rules are the mechanism. Without them the ten verticals are a form.

- **One to three questions per turn.** Never a dump. A form gets a form-shaped answer.
- **Probe every vague answer before advancing.** The vertical is not finished because a sentence was produced.
- **Record "I do not know" as a gap marker and continue.** It is a legitimate answer and it is data. Do not coach it away.
- **Never answer for them.** If you find yourself writing the persona, stop, and ask instead.
- **No padding.** Every sentence in the synthesis adds information or comes out.

| Vague answer | The probe |
| --- | --- |
| "Everyone who does X" | "Name the last three real people you have spoken to who match that" |
| "It saves them time" | "How many hours, doing what, and who currently gets paid for those hours" |
| "We are cheaper" | "Cheaper than what they use now, or cheaper than the competitor they never bought" |
| "It goes viral because it is useful" | "Name the artefact a non-user sees and where they see it" |
| "We will do content marketing" | "Which article, ranking for which phrase, read first by whom" |
| "Enterprises are interested" | "Which named role at which size of company, and what did they do after the call" |

## The decision rule

After the tenth vertical, count the gap markers and take exactly one branch.

- **Fewer than three gaps, none of them in verticals 1, 2 or 4.** Viable on the current plan. The report is a sequencing document rather than a warning.
- **Three to six gaps, but the offer, the audience of one and the use-case library are solid.** Launch is not blocked. The gaps are distribution and they are cheaper to fix with live users than in advance. Prioritise them and say so.
- **Any gap in verticals 1, 2 or 4.** Do not launch yet. Those three are load-bearing: an unclear offer, an unnamed buyer, or a use-case library with no instant-return case will not be repaired by more channels.
- **Every vertical answered fluently, with no person behind any answer.** This is the **you cannot tell** branch, and it is both the most common and the most dangerous state: an articulate founder with a complete document and zero evidence. Do not issue a verdict. Issue an evidence plan instead. Name the five claims the whole blueprint rests on, and for each one give the cheapest test and the number of conversations needed, with a bias towards past behaviour rather than stated intent. State plainly that what you produced is a well-formed hypothesis, not a validation, and that a verdict issued now would be a verdict on the founder's fluency.

## Phase 3: the report

Five sections, in this order, with these counts.

1. **Idea summary.** The confirmed play-back from phase one, two to three sentences, unchanged.
2. **The three strongest verticals**, with one line each on why.
3. **The three weakest verticals, stated bluntly.**
4. **The top five prioritised pre-launch actions.** Each one starts with a verb, names who does it, and states its completion test.
5. **The viability verdict.** The single biggest risk, and what would have to be true for this to work.

On bluntness. The instruction is not stylistic. A softened gap is read as an optional improvement and scheduled for after launch, which is the moment it stops being cheap to fix.

| Diplomatic, and useless | Blunt, and actionable |
| --- | --- |
| "The keyword strategy could benefit from further development" | "There is no keyword strategy. You named three head terms owned by companies with large content teams, and no long-tail set at all" |
| "Consider strengthening the growth loop" | "The loop as described cannot exist. The output is published under the customer's brand, so no non-user ever sees yours" |
| "Pricing may warrant revisiting" | "You are charging per seat into a market where the seats change every week. Revenue will fall in your customer's best months" |

Bluntness is about the claim, not the tone. Say the thing exactly once, with the evidence from the interview, and do not repeat it.

## Batch mode

When this runs non-interactively inside a larger pipeline, there is no founder to interview. In that case: skip phases one and two, produce only the phase three report from the stated idea, mark every inferred answer as inferred, and say in the first line that this is a batch run and roughly which verticals were answered by inference. A batch report that does not announce itself is the worst output this skill can produce, because it reads exactly like an assessment.

## Worked example, compressed

**The idea.** A tool that turns resolved support ticket threads into draft help-centre articles. Sold to support leads at software companies of twenty to two hundred people. Stage: private beta, six users.

**Play-back, confirmed.** It watches a resolved ticket thread, drafts a help-centre article from it, and puts the draft in front of a support lead for approval. It is for support leads whose team answers the same question repeatedly. Today they either write the article manually at the weekend or never write it.

**Vertical 1.** Offer sentence had no outcome in it. Rewritten to name the outcome, which is that repeat tickets fall. Pricing per agent seat. Risk removal: none. Gap marker.

**Vertical 2.** The person: a support lead, two staff, four products, judged on first-response time and repeat contact rate. Blocking belief: that automatically drafted documentation will be wrong in public.

**Vertical 4.** Nine use cases. Seven fit the template after rewriting. Two were adjectives and were deleted. Three are instant return, all of them versions of "you resolved this thread, here is the article, approve it in thirty seconds".

**Vertical 6.** The real alternative is not a competitor. It is nothing, plus a wiki nobody updates.

**Vertical 8.** Three head terms, all owned by large documentation vendors. No long-tail set. Gap marker.

**Vertical 9.** The claimed loop: customers publish the articles, readers see them, some become customers. Stages one to three name fine. Stage four fails. The articles publish on the customer's own help centre, under the customer's brand, with no attribution. Gap marker, recorded as "not a loop yet".

**Vertical 10.** No pre-launch mechanism. Activation is nine steps from signup to first draft article. Gap marker.

**The report.**

Strongest: the audience of one, the instant-return use cases, and the honest competitor picture.

Weakest, blunt: there is no growth loop, because nothing a non-user sees carries your name. There is no keyword strategy, only three terms you will not rank for. And activation is nine steps long for a product whose entire promise is that the first article appears in thirty seconds.

Five actions: cut activation from nine steps to three, ending on a drafted article. Add an opt-in attribution line to published articles and measure whether customers leave it on. Replace the three head terms with fifteen long-tail support-lead questions taken verbatim from community threads. Recruit eight design partners from the beta list as the pre-launch mechanism. Test a per-product price against the per-seat price with four of them.

**Verdict: not ready for a public launch in six weeks, and the reason is not the product.** The single biggest risk is that the loop as designed cannot exist, so every user must be bought or persuaded individually with no compounding, on a per-seat price in a market with small teams. For this to work, one of two things has to become true: either the attribution line survives contact with real customers, or you accept that this grows through paid and partnerships and price it accordingly.

## Failure modes

**Form dumping.** All the questions in one message. From the outside: the founder answers in fragments, reuses the same three sentences across four questions, and the whole vertical is done in ninety seconds. The report that follows is confident and hollow.

**Advancing on vagueness.** A sentence was produced, so the box is ticked. Recognisable later, because the report contains phrases like "growing teams" and "increased efficiency" that no one in the interview ever defined.

**Answering for the founder.** You supply the persona, the use case or the loop because you can see what it should be. Symptom: a fluent document in which the founder does not recognise their own product, and cannot defend any of it in a meeting.

**Diplomatic gap reporting.** The three weakest verticals appear as considerations. Symptom: everyone agrees the report was helpful, nothing in the plan changes, and the same three gaps are still there at launch.

**Loop hand-waving.** A cycle written with a trigger, an action and a reward, and no named artefact or new-user path. Symptom: the loop section reads well and nobody can say where a stranger would first see the product.

**Skipping the play-back.** The framework starts before the idea is confirmed. Symptom: a correction arriving at vertical seven that invalidates verticals one to six, usually phrased as "well, not exactly".

**Persona inflation.** Four personas with the same pains and different job titles, which flatters the market size and hides the fact that only one of them has budget.

**Instant-return blindness.** Every use case pays back in month three, and it passes unnoticed because each one is individually plausible. Shows up after launch as a trial that nobody finishes.

**Keyword aspiration.** Head terms recorded as a strategy because they describe the product accurately. Symptom: six months of content aimed at phrases whose first page is entirely large incumbents.

## What this skill does not do

- It does not talk to your buyers, so it cannot distinguish a market from your account of one. With thin evidence it will hand you an evidence plan instead of a verdict.
- It has no live data. Keyword rankability is estimated from the shape of a results page, not from an index, and a real keyword tool beats it on that vertical alone.
- It does not compute loop arithmetic. Branching factor and cycle time need a running product with attributed origins, and the deliverable here is the loop specification.
- It does not build a financial model. Nothing in the ten verticals tells you whether the unit economics close.
- Run non-interactively it produces a hypothesis document, because the interview is the mechanism rather than the packaging.
- It does not replace launching. It reduces the number of things you find out expensively. It does not tell you whether people will pay.