Ng Lawyer DB Build

NG Lawyer DB Build (Step 1)

What this skill does

This is Step 1 of Fei Gao’s “Nigeria Lawyer Network” workflow:

1. Build a Lawyer Database (this skill)
2. Send outreach emails and score by response (separate skill)
3. Sort & export for fast lookup (separate skill)

The objective is to create a database that allows fast matching by: City → Practice → Sub-specialty → Regulator → Firm size → Position → Score → Contact.

This skill focuses on data collection + structured classification with evidence links.

Inputs

• city: Lagos | Abuja
• practice: Construction | RealEstate | Labour | Tax | Criminal | IP
• n_per_segment: integer (target 3)

Output files

• lawyer_db_ng.xlsx (main database)
• sources.jsonl (evidence per record: url + snippet + timestamp)

Database schema (fixed column order)

The Excel must contain columns in this exact order:

1. Lawyer_UID
2. Country
3. City
4. Practice
5. Sub_Practice
6. Regulator_Tag
7. Firm_Name
8. Firm_Size
9. Position
10. Lawyer_Name
11. Email
12. Phone
13. LinkedIn
14. Website
15. Evidence_URL
16. Score_Total (blank in Step 1)
17. Score_ResponseSpeed (blank)
18. Score_Detail (blank)
19. Score_Pricing (blank)
20. Score_Cooperation (blank)
21. Risk_Flag (blank)
22. NeedFollowUp (0/1)
23. ChannelCostRisk (Low/Medium/High)
24. Last_Updated

Non-negotiable rules (to avoid future scoring chaos)

1) Unique key (Lawyer_UID)

All later scoring MUST be written back by Lawyer_UID, never by name.

Format:

• LAW-NG-{CITYCODE}-{PRACTICECODE}-{NNNN}
• CityCode: Lagos=LAG, Abuja=ABJ
• PracticeCode: Construction=CON, RealEstate=REA, Labour=LAB, Tax=TAX, Criminal=CRI, IP=IPR

Example:

• LAW-NG-LAG-CON-0001

2) Email policy (anti-channel-cost)

• Email must be explicitly published on the evidence page (law firm bio page, author page, or official profile).
• NO guessing (e.g., firstname.lastname@firm.com is not allowed unless shown).
• If only a general firm email exists (info@ / contact@), it may be used but set:
  • ChannelCostRisk=High
  • NeedFollowUp=1

3) Evidence chain

Each row must store Evidence_URL. The skill also writes sources.jsonl with:

• Lawyer_UID
• Evidence_URL
• Evidence_Snippet (<= 300 chars)
• Captured_At

4) Firm_Size classification (fee proxy)

Firm_Size affects pricing expectation. Classify using evidence:

• Top:
  • The firm is ranked in Chambers or Legal 500 in Nigeria for the relevant practice; OR
  • Firm lawyer count > 50 (evidence required)
• Mid: lawyer count 15–50
• Small: lawyer count 2–14
• Solo: sole practitioner / independent (count=1)
• Unknown: insufficient evidence → set NeedFollowUp=1

5) Position mapping (fee proxy)

Normalize to:

• Partner / Counsel / SeniorAssociate / Associate / Junior / Independent / Unknown

Mapping hints:

• Partner: Managing Partner, Senior Partner, Partner
• Counsel: Counsel, Of Counsel
• SeniorAssociate: Senior Associate, Associate (Senior)
• Associate: Associate, Solicitor
• Junior: Junior Associate, Trainee
• Independent: Sole Practitioner, Principal, Independent Legal Practitioner

Regulator_Tag (only with evidence)

Allowed values:

• CAC, NIPC, FIRS, StateIRS, NAFDAC, Immigration, MinistryOfMines, IPRegistry, EFCC, Police, Court

Rule:

• Tag only when evidence mentions regulator or shows relevant matters.
• If not evidenced, leave blank; outreach email will ask the lawyer to self-report regulators handled.

Data collection strategy (recommended sources)

Priority order:

1. Chambers / Legal 500 ranked firms for Nigeria practice (Top candidates)
2. Official law firm team pages (bio pages with emails)
3. Author pages on firm sites / Lexology / IFLR (when email is shown)
4. Boutique firm sites for Small/Solo

Segmentation target

For each (Firm_Size x Position) segment, collect n_per_segment candidates with verified emails where possible.

Recommended priority within Top firms:

• Counsel / SeniorAssociate / Associate (often more willing for flexible cooperation)

Example run (MVP)

City: Lagos
Practice: Construction
n_per_segment: 3

Goal: produce a usable initial database for Lagos Construction to validate:

• Firm_Size classification is correct
• Position mapping is correct
• Emails are evidence-backed
• Column order is stable