Industry Research Production

This drill produces a single data-driven industry report from scratch: topic validation, data collection, analysis, writing, and publication. The output is a hosted, gated or ungated report ready for distribution. Designed for founder-led content plays targeting problem-aware audiences.

Input

• ICP definition (from icp-definition drill): target titles, company sizes, pain points
• Founder's domain expertise and any proprietary data sources (product usage data, customer conversations, internal benchmarks)
• Competitive landscape: what reports competitors have already published
• Distribution plan: how and where the report will be promoted (feeds into report-distribution-pipeline drill)

Steps

1. Identify the report topic

Use the news-signal-search fundamental to scan for trending industry topics in the last 90 days. Query news APIs and social listening for:

• Pain points your ICP mentions repeatedly on LinkedIn, Reddit, and industry forums
• Emerging trends without authoritative data (the gap you can fill)
• Regulatory, market, or technology shifts creating uncertainty

Use the competitive-intel-aggregation fundamental to audit existing reports from competitors and analysts:

• Search for "{industry} + report/benchmark/survey + {year}" on Google
• Document: who published it, what data they used, what conclusions they drew, what gaps remain
• Prioritize topics where existing reports are >12 months old, rely on anecdotal evidence, or miss your ICP's specific context

Select a topic that meets all three criteria: (a) your ICP actively cares about it, (b) no authoritative recent data exists, (c) you can collect original data.

2. Design the data collection methodology

Choose one or more data sources based on what you can access:

Option A -- Survey-based (best for benchmarks and "state of" reports): Use the typeform-survey-setup or tally-form-setup fundamental to build a 10-15 question survey. Keep it under 5 minutes to complete. Structure:

• 3 demographic/firmographic questions (company size, role, industry)
• 5-8 quantitative questions (Likert scales, numeric ranges, multiple choice)
• 2-3 open-ended questions (qualitative insights for quotes)

Target: minimum 50 responses for statistical relevance. For niche B2B, 30-50 is acceptable if the respondent profile is well-defined.

Option B -- Enrichment-based (best for landscape and market maps): Use the clay-claygent fundamental to research 100-200 companies matching your ICP. For each, ask Claygent structured questions about their approach to {topic}: tooling, team size, process maturity, public statements. Aggregate into quantitative patterns.

Use the clay-enrichment-waterfall fundamental to pull firmographic data (headcount, funding, tech stack) for each company. Cross-reference with the qualitative findings to identify correlations (e.g., "companies with >100 employees are 3x more likely to have a dedicated {function}").

Option C -- Product data (best for benchmark reports): If your product generates usage data relevant to the topic, anonymize and aggregate it. Example: "We analyzed {N} accounts over {time period} and found {insight}." Even 20-50 data points create credibility when the methodology is transparent.

Option D -- Public data scraping (best for market landscape reports): Use clay-claygent to pull data from public sources: job postings mentioning relevant skills/tools, G2/Capterra review trends, GitHub stars, app store rankings, or SEC filings. Structure the queries to produce comparable data across companies.

3. Collect and clean the data

Execute the collection method chosen in step 2:

For surveys:

• Distribute via email to existing contacts, LinkedIn posts, and relevant communities
• Use the loops-broadcasts fundamental to email your subscriber list with the survey link
• Offer incentive: early access to the report or an executive summary
• Run collection for 2-3 weeks minimum
• Clean data: remove incomplete responses, flag outliers, verify no duplicate submissions

For enrichment/scraping:

• Run Clay tables and verify output quality on first 10 rows before scaling
• Flag data points that seem anomalous and verify manually
• Structure all data in a consistent schema (Google Sheets or CSV)

4. Analyze the data

Load cleaned data into Google Sheets or a pandas notebook. Produce:

• Summary statistics: mean, median, distribution for each quantitative variable
• Cross-tabulations: break metrics by company size, industry, maturity level
• Key findings: 5-8 data-backed insights that surprised you or contradict conventional wisdom
• Quotable statistics: 3-5 headline numbers that are shareable on social media (e.g., "72% of B2B teams still do X manually")
• Visualizations: 4-6 charts (bar charts, pie charts, scatter plots) suitable for embedding in the report and extracting as social graphics

Use the Anthropic API to assist with analysis:

POST https://api.anthropic.com/v1/messages

System: "You are a data analyst reviewing survey/research results for an industry report. Identify the 5 most surprising or actionable findings. For each, provide: the specific data point, why it matters to {ICP_DESCRIPTION}, and a one-sentence headline suitable for social media."

User: "Here is the raw data summary: {DATA_SUMMARY}. Here are the cross-tabulations: {CROSS_TABS}."

5. Write the report

Using the ai-content-ghostwriting fundamental, generate the report via the Anthropic API:

POST https://api.anthropic.com/v1/messages

System: "You are writing an industry report authored by {FOUNDER_NAME}, founder of {COMPANY}. The report targets {ICP_DESCRIPTION} who are problem-aware about {PAIN_POINT}. Write in the founder's voice: {VOICE_PROFILE}. Requirements:
- Executive summary (200 words): headline findings and why they matter
- Methodology section (100 words): how data was collected, sample size, limitations
- 5-8 findings sections: each with a headline stat, supporting data, chart reference, and actionable recommendation
- Conclusion: what this means for the reader's business
- Soft CTA: single mention of {PRODUCT} as relevant context, not a pitch
- Total length: 2,000-4,000 words depending on findings depth"

User: "Key findings: {FINDINGS}. Raw data summary: {DATA_SUMMARY}. Charts available: {CHART_LIST}. Competing reports and their gaps: {COMPETITOR_ANALYSIS}."

Human action required: Founder reviews the full draft. Key checks:

• Are all data points accurate and sourced correctly?
• Do the conclusions follow from the data (not AI hallucination)?
• Does the report contain at least one genuinely surprising insight?
• Does it read as the founder's perspective, not generic analyst language?
• Remove any hedging language ("it appears that", "one might consider") -- be direct

6. Design and publish

Format the report:

• PDF version: Clean, branded layout. Include charts inline. Add a cover page with title, author, and date. No heavy design -- founder-authored quality, not agency-produced.
• Web version: Using the ghost-blog-publishing fundamental, publish a landing page version with gated or ungated access. If gated, keep the form minimal (email + company name). If ungated, add an email capture CTA for "future reports."
• Social graphics: Extract 4-6 charts and quotable stats as standalone images sized for LinkedIn (1200x627) and Twitter (1600x900).

Host at a memorable URL: {yourdomain}.com/reports/{report-slug}

Configure PostHog tracking on the landing page:

• report_page_viewed -- properties: report_slug, source, utm params
• report_downloaded -- properties: report_slug, email (if gated)
• report_shared -- properties: report_slug, platform

Output

• One published industry report (PDF + web version)
• 4-6 social graphics with quotable stats
• Landing page with PostHog tracking configured
• Raw data set for future analysis and report updates
• Distribution brief: key findings, shareable quotes, and suggested social posts (feeds into report-distribution-pipeline)

Triggers

Run this drill once per quarter at Scalable/Durable levels. At Smoke level, run once to validate the concept. At Baseline, run 1-2 times over 30 days. Total production time per report: 8-15 hours depending on data collection method.