Export CSV

Convert pipeline JSON files to CSV. Each file gets its own CSV saved in the same folder.

Available exports

What to ask

If the founder does not specify which file(s) to export, ask:

Which files would you like to export to CSV?

  1. All available files
  2. fund_candidates — candidate VC firms
  3. investor_list — mapped people at each firm
  4. investor_contacts — enriched contacts with emails
  5. scored_funds — ranked firms with score breakdown
  6. final_output — full outreach list with drafted messages
  7. company_profile — your fundraising brief

Reply with a number, a comma-separated list (e.g. "3,4"), or "all".

Only export files that exist. Skip and note any that are missing.

CSV column definitions

fund_candidates.csv

fund_id, fund_name, website, location, stage_focus, sector_focus,
check_size_min_usd, check_size_max_usd, lead_behavior,
recent_investments, preliminary_rationale, possible_concerns, evidence_confidence

• stage_focus and sector_focus: join array items with | (e.g. seed | series-a)
• recent_investments: join array with |
• check_size_range_usd[0] → check_size_min_usd, [1] → check_size_max_usd

investor_list.csv

fund_id, name, title, firm, location, linkedin, firm_profile_url, why_relevant

investor_contacts.csv

fund_id, name, title, firm, location, linkedin, firm_profile_url,
work_email, email_confidence, source_type, contact_tier, reason_to_contact

scored_funds.csv

Flatten score_breakdown into individual columns:

fund_id, fund_name, score_total, score_max_possible,
score_sector_fit, score_stage_fit, score_geography_fit,
score_check_size_fit, score_lead_behavior_fit, score_portfolio_adjacency,
score_recency, score_partner_relevance,
score_notes, why_match, risks

• score_breakdown.sector_fit → score_sector_fit, etc.
• why_match: join array with |
• risks: join array with |
• Do not include the nested investors array — use investor_contacts.csv for that

final_output.csv

priority_rank, fund_name, score_total, score_max_possible,
investor_name, investor_title, contact_tier, work_email, linkedin,
outreach_subject_line, outreach_email_body, outreach_linkedin_dm,
outreach_personalization_hook, outreach_cta

This is the same as output.csv written at the end of step 6.

company_profile.csv

Single row. Flatten the object:

company_name, one_line_thesis, sector, stage, geo, business_model,
check_size_min_usd, check_size_max_usd, total_round_target_usd,
traction_summary, comparable_companies,
investor_lead, investor_thesis_keywords, investor_avoid, warm_intro_preferred,
assumptions_made, open_questions

• Arrays: join with |
• check_size_target_usd[0] → check_size_min_usd, [1] → check_size_max_usd
• investor_preferences.lead → investor_lead, etc.

How to write the CSV

Always write the Python conversion script to a .py file first, then execute it. Never use a bash heredoc (<<'EOF') to run Python inline — it triggers a false-positive safety warning due to curly braces in dict comprehensions.

Correct pattern:

1. Write the script to scripts/export_csv.py using the file-write tool
2. Run it: python3 scripts/export_csv.py
3. Delete the script after it runs (optional)

Wrong pattern (do not use):

python3 - <<'EOF'
...script with {curly braces}...
EOF

CSV formatting rules

• Always include a header row
• Wrap any field containing commas, newlines, or double quotes in double quotes
• Escape internal double quotes as ""
• Use UTF-8 encoding
• Null values → empty cell (not the string "null")

After export

List each file written with its row count:

Exported:
  ✓ data/raw/fund_candidates.csv — 24 rows
  ✓ data/normalized/investor_contacts.csv — 67 rows
  ✓ data/normalized/final_output.csv — 67 rows

Skipped (file not found):
  - data/normalized/scored_funds.json