# Demand Research

> Tests whether a proposed product bet has credible demand. Use when Why now depends on an unresolved, decision-critical demand claim or when the user explicitly asks for demand research. Do NOT use for child features, mandated work, parity work, or when a cheaper experiment can answer the question before research.

- Skill: `arcadeai/demand-research` (Agent Skill)
- Install (CLI): `npx skillmds@latest add arcadeai/demand-research`
- Raw SKILL.md: https://api.skillmd.com/api/skills/arcadeai/demand-research/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Research & Search
- Author: ArcadeAI (https://skillmd.com/u/arcadeai)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/arcadeai/demand-research

---


# Demand Research

Answer one question: **is there enough credible demand evidence to make this
product bet now?** Keep the result small enough to paste into Product Bet.

## First decide whether research earns its cost

Skip research when any applies:

- The ticket is a child contribution; inherit the epic's demand case.
- The work is mandated, compliance-driven, or parity work.
- A cheap reversible experiment can test the assumption more directly.
- The demand claim would not change scope, priority, or the build/no-build call.

When skipping, state the reason in one sentence. Do not manufacture a research
appendix to make the plan look complete.

## Research loop

1. Frame the single decision neutrally so either outcome is acceptable, and name
   the demand claim that could change it. Do not frame the task as proving demand.
2. Start with first-party evidence: customer requests, usage, sales/support
   records, interviews, or the user's supplied material.
3. Fill only meaningful gaps with current external sources. Prefer primary
   sources and record the date checked.
4. Grade evidence by sacrifice: behavior, money, time, or switching cost beats
   stated interest. Separate direct evidence from inference.
5. Run a disconfirmation pass: seek the strongest evidence against demand,
   including non-adoption, abandonment, unwillingness to pay, and contradictory
   usage. Report supporting and contradicting evidence together, name the
   strongest alternative explanation, and say what would change the verdict.
   When the evidence permits competing interpretations, state the ambiguity
   instead of resolving it in the sponsor's favor. Missing negative evidence is
   unknown, not support.
6. Stop when the decision can be made or the cheapest next validation is clear.

## Output

Return exactly these four compact parts:

- **Verdict — PRESENT, WEAK, ABSENT, or UNAVAILABLE:** one sentence tied to the
  decision, not a market-size claim.
- **Strongest evidence:** up to three dated bullets with links or local evidence
  identities; label inferences.
- **Gaps:** only unknowns that could reverse the verdict.
- **Cheapest validation:** one concrete next test, or `none` when evidence is
  already sufficient.

`ABSENT` and `UNAVAILABLE` are evidence states, not approval blockers. Do not
invent evidence or upgrade demand strength when sources are unavailable.

Update **Problem / Why now** with the verdict and strongest decision-bearing
evidence. Do not add a separate research document unless the evidence itself is
too large to cite compactly.