Market Research skill for Claude Code

Done well means: every number has a source, a date, and a confidence label; sizes are computed two independent ways and the gap is explained; assumptions are listed where a reader can change them; and the brief ends with a decision-relevant answer, not a pile of statistics.

Hard rules

• Never invent a statistic, report title, or analyst figure. If a number cannot be sourced, derive it from stated assumptions and label it estimate.
• Cite the original publisher of each figure, not the blog that repeated it. Trace “according to a report” chains back to the source or drop the figure.
• Record the year and geography of every figure; a 2019 global number is not a 2025 US number.
• Treat paid-report teaser numbers (press releases from market-research vendors) as low confidence: methods undisclosed, often inflated CAGRs. Never use them as the only basis for a size.
• Separate what you found from what you infer. Mark each claim sourced, derived, or assumption.

Workflow

- [ ] 1. Pin the decision and the market definition
- [ ] 2. Segment the customers
- [ ] 3. Size bottom-up (units x price)
- [ ] 4. Size top-down (category spend x share) and reconcile
- [ ] 5. Gather demand, competition, and pricing signals
- [ ] 6. Stress-test assumptions (sensitivity)
- [ ] 7. Write the brief with sources and confidence labels

1. Define the market

Write before searching:

• The decision this research informs (enter / pitch / price / prioritize a segment).
• The product and the job it does; the buyer (who pays) vs. the user.
• Geography, time frame, and currency (state nominal vs. real).
• What is in and out of scope: adjacent categories, substitutes, DIY/manual alternatives.

2. Segment

Segment by variables that change buying behavior, not demographics by default:

B2B	B2C
Company size (employees, revenue), industry, tech stack, regulatory regime, buying center	Need state / job-to-be-done, usage frequency, willingness to pay, channel, life stage

For each segment: count of potential buyers, pain intensity, current alternative, reachability (channel), and price tolerance. A good segment is identifiable, reachable, and large enough to matter.

3-4. Size the market

Definitions:

• TAM: total annual revenue if every potential buyer in the defined market bought.
• SAM: the part of TAM your product and go-to-market can actually serve (geography, segment, channel, product fit).
• SOM: what you can realistically capture in 3-5 years given competition and capacity. Justify with comparable companies’ penetration, not a flat “1%”.

Bottom-up (preferred, auditable):

Buyers in segment       = 48,000 US dental practices (source: census/industry association, year)
x Adoption ceiling      = 60% (assumption: share with digital scheduling, from survey X)
x Annual price          = $3,600 (derived: $300/mo median of 6 competitor price pages, 2025-09)
= SAM                   = $103.7M

Top-down (sanity check):

Category spend (practice-management software, US)  -> source A
x Share attributable to the scheduling job          -> assumption, justified

Reconcile: if the two differ by more than ~2x, find the bad assumption (usually buyer count, adoption, or price) rather than averaging. Report a range (low/base/high), not a single point.

Value-theory sizing for new categories: buyers x value created x capturable share (typically 10-30% of value), when no category spend exists yet.

5. Signals and desk research

Prefer primary data over aggregators. Source types by reliability:

Tier	Source	Use for
A	Government statistics (Census, BLS, Eurostat, national statistics offices), regulators, central banks	Counts of businesses/households, spend, employment, prices
A	Company filings (SEC EDGAR 10-K/10-Q/S-1, annual reports), earnings call transcripts	Revenues, customer counts, segment sizes, pricing
B	Industry associations, academic studies, reputable surveys with published methods	Adoption, behavior, willingness to pay
B	Public pricing pages, job postings, app-store data, search volume trends	Pricing, demand momentum, competitor focus
C	Analyst press releases, vendor-sponsored reports, news articles	Leads to check, never the sole source

Demand signals worth gathering: search volume trends, job postings mentioning the problem, community discussion volume, competitor funding and hiring, public customer counts, review counts on marketplaces. For keyword volumes, the keyword-research skill covers tools and methods. For a competitive landscape, use competitor-analysis.

For a directory of free data sources by region and topic, read references/data-sources.md when you need specific numbers.

6. Stress-test

• List every assumption with its range; vary each to low/high and record the effect on SAM. Lead the brief with the 2-3 assumptions that move the answer most.
• Check for double counting across segments and for TAM that includes buyers who cannot legally or practically buy.
• Ask what would have to be true for the high case, and whether any evidence supports it.

Confidence labels

Label	Meaning
High	Tier A source, recent (<= 2 years), matches definition and geography
Medium	Tier B source, or Tier A needing adjustment (older, broader scope)
Low	Tier C source, single unverified figure, or a key assumption without data

Output template

# Market brief: <market> (<geography>, <year>)
**Decision supported:** ...   **Bottom line:** SAM $X-Y M (base $Z M); attractive / marginal / unattractive because ...

## Market definition and scope
## Segments (table: segment, buyer count, pain, alternative, price tolerance, priority)
## Sizing
| Metric | Low | Base | High | Method | Key sources |
|--------|-----|------|------|--------|-------------|
| TAM | | | | top-down | |
| SAM | | | | bottom-up | |
| SOM (yr 3) | | | | comparable penetration | |
Assumptions table: assumption, value, range, basis (sourced/derived/assumption), confidence
## Demand and competition signals
## Pricing benchmarks (source URL + access date per price point)
## Risks and open questions (what primary research would resolve them)
## Sources (publisher, title, year, URL, access date)

Gotchas

• CAGR figures from market-research press releases routinely conflict by 3-5x for the “same” market; definitions differ. Do not quote one without its definition.
• TAM framed as “global spend on <broad category>” impresses no investor; a credible bottom-up SAM does.
• Revenue and spend are not the same as units; check whether a source counts dollars, seats, companies, or installs.
• Pricing pages change; record the access date and archive the page (web.archive.org) when the number matters.
• Survey-based willingness to pay overstates real purchase behavior; discount it or validate with price tests.