company-deep-dive-report

Company Deep-Dive Report

Do not summarize the company; elucidate the causal structure that drives decisions. Connect every important claim to evidence, and separate fact, inference, hypothesis, and unknown.

Read the necessary reference files

• At the start of analysis, read references/analysis-routing.md and determine the company type and analysis route.
• Before collecting information, read references/evidence-contract.md and fix the evidence ledger and claim standards.
• When designing deliverables, read references/report-architecture.md and determine the page structure and visual rules.
• assets/ Copy the template into the execution folder and use it. Do not edit the template itself.
• If you want to check the completed data structure, refer to examples/sample_run/ . Since this is a fictional company, do not reuse the analysis content for a real company.

1. Fix the analysis conditions

1. Identify the target company by corporate name, country, official URL, ticker, etc., and distinguish companies with the same name.
2. Turn the following items into a one-page analysis brief.
   • Readers and decisions
   • Central question
   • Target period and region
   • Comparison targets
   • Desired deliverable format
3. Fill minor gaps with the following default values and state them at the beginning.
   • Readers: Management and business strategy staff
   • Central question: What is the structure that creates this company's value? Will that structure strengthen over the next 1–3 years?
   • Deliverables: A 12-page-equivalent Markdown report, evidence ledger, claim graph
4. Obtain confirmation before using paid information, information requiring authentication, personal information, or collection methods that may violate terms of service.
5. By default, make company-deep-dive/<company-slug>/<YYYY-MM-DD>/ the execution folder.

2. Classify the company and choose the research route

1. Classify listed/unlisted, single-business/conglomerate, growth/mature, and information volume.
2. Among SaaS, Marketplace, manufacturing, consumer goods/retail, finance, media, infrastructure, and professional services, choose the main business type.
3. Switch KPIs, primary sources, competitive units, and risk items according to the classification result.
4. Leave the basis for classification and uncertainties in analysis_brief.yaml .
5. Do not apply generic frameworks first. First hypothesize the company-specific value creation mechanism.

3. Design hypotheses and evidence slots

1. Narrow down to 3–5 hypotheses that drive the final decision.
2. For each hypothesis, define the following.
   • Implications if supported
   • Meaning if disproven
   • Required evidence
   • Leading and lagging indicators
   • Disproof conditions
3. Convert each page of the report into evidence slots to be filled.
4. Search not only for supporting evidence but also for disproof, contradictions, missing data, business withdrawals, and changes in metric definitions.

4. Collect and ledger evidence

1. For corporate information that may change, use the Web, connected data, and corporate disclosure materials available at that point in time.
2. Prioritize primary sources such as company disclosures, regulatory authorities, statistics, patents, and litigation and administrative materials.
3. Treat news, industry materials, reviews, and social media as auxiliary signals, and do not use them alone to assert performance or market size.
4. Give each piece of evidence an ID in the format E001 and record it inevidence_ledger.csv .
5. Always attach units, periods, regions, definitions, and sources to numerical values.
6. Do not fill in information that cannot be obtained with speculation; leave it asUNKNOWN and note the additional research required.

5. Build the company's causal model

Analyze in the following order.

1. Create a business model diagram connecting customers, value provided, monetization, supply, and channels.
2. Break down the path from revenue to profit and cash into a company-specific KPI driver tree.
3. Compare economics and resource allocation by segment, region, and customer group.
4. Show competitive advantage through the causal chain of "assets and capabilities → customer behavior → economic effects."
5. For each growth driver, evaluate headroom, execution capability, evidence, and constraints.
6. Register each claim in the format C001 in claim_graph.json , and connect it to evidence IDs.

Do not stop at merely listing SWOT, 3C, and PEST. For each diagram, write "so what happens to performance, competition, and decision-making?"

6. Pass through Evidence Court

1. Do not count reprinted information or information from the same announcement as multiple independent sources.
2. Classify each claim as FACT、INFERENCE、HYPOTHESIS、UNKNOWN .
3. For each important claim, check disproof evidence, alternative explanations, and disproof conditions.
4. Do not forcibly integrate contradictory information; leave both sides asCONTRADICTION .
5. Indicate confidence as HIGH、MEDIUM、LOW , and write the reason in one sentence.
6. Tier 1 important claims generally require at least two independent pieces of evidence, of which at least one must be primary information.

If the conditions cannot be met, weaken the assertion and clearly state the missing evidence as a partial report.

7. Create future scenarios

1. Define three scenarios—Bull, Base, and Bear—not as narratives but as changes in causality.
2. Attach the following to each scenario.
   • Trigger conditions
   • Propagation path to KPIs
   • Leading indicators
   • Disproof conditions
   • Strategic implications
3. Make the 5 to 10 indicators that should be monitored in the near term into a Strategic Watchlist.
4. When placing forecast values, show the assumptions, calculation formula, range, and source.

8. Generate deliverables

At a minimum, create the following.

• company_dossier.md: A corporate report equivalent to 12 pages
• evidence_ledger.csv: Evidence ledger
• claim_graph.json: Connections between claims and evidence
• analysis_brief.yaml: Analysis conditions and classification results

If tools for creating slides, Word, or PDF are available, add editable deliverables while maintaining the same claim IDs and evidence IDs. Always render the materials and visually check them.

9. Inspect completion conditions

Do not consider it complete until all of the following are satisfied.

• The Executive Thesis directly answers the central question.
• The business model diagram and KPI tree are specific to the target company.
• Important claims can be traced to evidence IDs.
• Units, periods, regions, and definitions for numerical values are not missing.
• Facts, inferences, hypotheses, and unknowns can be distinguished.
• Not only supporting evidence but also counterevidence and contradictions are addressed.
• The 3 scenarios have leading indicators for judgment.
• Each page of the report is focused on a single main message.

Finally, execute the following and fix any errors.

python <skill-directory>/scripts/validate_run.py <run-directory>

In the execution environment, python If cannot be used, manually check the same inspection items and communicate that they were not executed.