biz-market-analysis

Mission

Produce market intelligence and customer insights that are accurate, verifiable, traceable, and action-oriented. Treat every important conclusion as a claim that needs evidence, source metadata, methodology notes, limitations, and confidence.

This skill applies to both consumer products, such as an AI smart thermos, and industry/B2B products, such as a green-building integrated energy intelligent control system.

Default final output is exactly three files:

1. market_customer_insight_sources.xlsx: searched data sources, source metadata, extracted fields, evidence records, and DVC validation results.
2. market_customer_insight_dashboard.html: a single-page visual dashboard for market information and customer insight data.
3. market_customer_insight_report.doc: a formal Word-compatible report for structured explanation and review.

Do not create Markdown, JSON, PPT, PNG/SVG chart packs, or extra final files unless the user explicitly changes this constraint.

Operating Principles

1. Start from the decision, not from search keywords: identify what the user needs to decide, the target market, geography, customer segment, and time horizon.
2. Decompose the task into research questions, hypotheses, indicators, and source plans before forming conclusions.
3. Prefer high-quality primary or authoritative sources: official statistics, regulatory filings, company financials, prospectuses, standards, internal first-party customer data, interviews, CRM, support tickets, and transaction data.
4. Separate facts, estimates, inferences, hypotheses, and recommendations. Never present an unverified hypothesis as fact.
5. Every key claim needs an evidence trail whenever possible: source, URL or file reference, date, data period, methodology, metric, evidence level, quality score, limitation, and confidence.
6. Triangulate major claims with multiple source types. Do not let PR, news, social buzz, fundraising, or a single market-report summary prove demand by itself.
7. Translate customer insight into action: target segment, use case, job-to-be-done, pain, trigger, objection, alternative, willingness to pay, product implication, GTM message, and validation experiment.
8. Flag uncertainty directly. If evidence is weak, label the result as "early signal", "assumption", or "to validate".
9. Protect privacy: anonymize and aggregate customer data; avoid exposing personal sensitive information.

Reference Loading

Load only the reference files needed for the task:

• Read references/evidence-engine.md when collecting, scoring, validating, or citing evidence.
• Read references/deliverables.md when generating the required three files: Excel source results, HTML dashboard, and Word .doc report.
• Read references/industry-playbooks.md when adapting the workflow to consumer goods, B2B software, industrial systems, smart hardware, regulated markets, or investment/strategy work.
• Read references/multi-agent-platforms.md when designing or running a multi-agent workflow, including OpenClaw, WorkBuddy, or similar agent platforms.

Use scripts/research_package.py to initialize or validate a traceable research package.

Core Workflow

1. Intake and scope

   • Extract product or industry, user decision, geography, target customer, value proposition, competitors, available internal data, and required deliverables.
   • Ask at most three clarifying questions only if missing information would materially change the analysis. Otherwise state assumptions and proceed.

2. Research plan

   • Use the 6D structure to plan the research:
     • D1 Define: decision, audience, scope, assumptions, success criteria.
     • D2 Decompose: market/customer/competitor/policy/technology dimensions and subquestions.
     • D3 Design Fields: fields that must be collected for each source and analysis object.
     • D4 Discover: source channels, keywords, search order, source exclusion rules.
     • D5 Digest: dedupe, clean, verify, resolve conflicts, cite.
     • D6 Deliver: exactly three final files and their acceptance criteria.
   • Define market boundaries and adjacent categories.
   • Generate research questions and hypotheses.
   • Map each hypothesis to measurable indicators and candidate data sources.
   • Decide whether the task is Lite, Standard, or Deep Research.

3. Evidence plan and collection

   • Design the deliverables before designing search queries.
   • Search from higher-trust sources first, then use lower-trust sources only to discover signals, customer language, and competitor claims.
   • Use Chinese and English keywords when the market, technology, or competitor landscape is cross-border.
   • Prioritize A/B/C-level sources for market size, growth, regulatory, and company claims.
   • Use D/E-level sources for competitor self-description, pricing, messaging, early demand signals, and customer pain discovery, with limitations noted.
   • For recent market facts, live prices, current regulations, rankings, or active competitors, verify with current sources and cite them.

4. DVC validation and evidence ledger

   • Apply DVC at field or claim level: locate original source, verify fields, check validity period, assign DVC grade, and provide adoption guidance.
   • Record dvc_grade as A/B/C/D and adoption_recommendation as adopt/reference/do_not_use_yet/exclude.
   • If a conclusion depends on multiple key fields, the conclusion cannot exceed the lowest critical DVC grade.
   • Convert sources into evidence objects.
   • Score evidence quality and assign A/B/C/D/E level.
   • Link every important claim to source_id and evidence_id.
   • Display conflicts and explain possible reasons such as geography, period, sample, channel, or methodology differences.

5. Analysis

   • Market: define category, estimate TAM/SAM/SOM or relevant market size from bottom-up variables where possible, analyze growth drivers and constraints.
   • Competitors: classify direct competitors, indirect competitors, substitutes, and potential entrants; compare positioning, features, pricing, channels, customer feedback, and changes over time.
   • Customers: cluster segments by task and buying context, not only demographics; identify pains, triggers, objections, alternatives, buying committee, and willingness to pay.
   • Opportunity: score opportunity by demand intensity, evidence quality, willingness to pay, competitive intensity, feasibility, timing, regulatory risk, and GTM accessibility.

6. Deliver

   • Always include an executive conclusion, scope and assumptions, key evidence table, uncertainties, and next validation plan.
   • Prefer tables and charts when data supports them.
   • Produce only the three required final files: Excel source results, HTML single-page dashboard, and Word .doc report.
   • Make the HTML dashboard and Word report cite the same source_id, evidence_id, and DVC fields stored in the Excel file.

Quality Gates

Before finalizing, verify:

• Key claims have evidence IDs or are explicitly marked as assumptions.
• Search followed 6D: D1-D6 are represented in the source workbook or report method section.
• Each key data point has DVC field verification and adoption guidance.
• Market-size claims do not rely only on news, PR, or one report abstract.
• Customer pain claims use customer-side evidence where available.
• Competitor claims from official websites are labeled as competitor self-description.
• Charts are tied to source data and do not compare incompatible methodologies.
• Recommendations are specific enough to guide product, sales, marketing, strategy, or validation experiments.
• Output includes risks, limitations, and what data would change the conclusion.
• The final deliverable folder contains exactly the three expected files, unless the user explicitly requested otherwise.

If the evidence is insufficient, do not stretch. Return a defensible partial conclusion and a prioritized data collection plan.