mcsqstudio/africa-startup-directory

MC Studio Africa Startup Directory

Dataset card for the curated company directory built by MC Studio (Nairobi-based venture studio, Silver IBM Business Partner) for its startup curation and investment-tracking work.

Summary

1,475 technology, agri-business and SME companies headquartered in Africa (plus African-founded companies operating from outside the continent). Each record combines a natural-language company_description with structured fields — founding year, company age, HQ country/city/region, employee band, operating countries, products & services — and two curated labels:

• aikya_tier — venture-support tier assigned per the Aikya Legal (Okutta & Wairi Advocates) model: HUSTLE → GROW → LEAD, or Unknown.
• operational_sector — Sita Sector Program operational-sector code derived by sieving the company description against per-sector trigger-word lists.

This is tabular/text data intended for sector and tier classification experiments, market intelligence, and pipeline triage.

Collection & curation

• Source: public company directories and websites compiled by MC Studio, 2026.
• Processing: raw records were combined, de-duplicated (unique company_id, e.g. COMP-000001), cleaned, validated, standardized and enriched with region mapping and sector sieving against per-sector trigger-word lists.
• The canonical export is data.parquet, derived from the Standardized - Model Data sheet of the source workbook.

Data structure

Column	Type	Description
company_id	string	Unique identifier (e.g. COMP-000001)
company_name	string	Company legal/common name
company_description	string	One-paragraph natural-language description
year_founded	int32	Founding year (null for 1 record)
company_age	int32	Age in years as of 2026 (null for 1 record)
hq_country	string	Headquarters country
hq_city	string	Headquarters city (null for 1 record)
hq_region	string	African region of HQ (6 values incl. Outside Africa)
operating_countries	string	Countries of operation (null for 7 records)
products_and_services	string	Semicolon-separated product/service list
employee_count	string	Banded headcount (e.g. 11–50)
aikya_tier	string	AIKYA HUSTLE, AIKYA GROW, AIKYA LEAD, Unknown
operational_sector	string	Sector code (see below)

Sector code legend (operational_sector)

Code	Meaning
ATX	Agritech
ETX	Edtech
HTX	Healthtech
FTX	Fintech
REC	Retail & e-commerce
ERG	Energy
MFG	Manufacturing
XSC	Cross-sector / multi-sector technology
Unknown	Not classified

These are the six Sita Sector Program priority industries (with XSC for cross-sector technology plays). The full trigger-word lists and country→region mapping used during labeling are maintained locally alongside the source workbook.

Distributions (n = 1,475)

Aikya tier: AIKYA LEAD 544 · AIKYA GROW 526 · AIKYA HUSTLE 306 · Unknown 99

Operational sector: XSC 385 · ETX 196 · ATX 195 · FTX 191 · HTX 189 · REC 152 · ERG 101 · MFG 63 · Unknown 3

HQ region: East Africa 516 · Outside Africa 347 · West Africa 270 · North Africa 200 · Central/Middle Africa 87 · Southern Africa 55

Top HQ countries: Kenya 176 · South Africa 155 · Nigeria 123 · Egypt 107 · Uganda 74 · Ghana 68 · Rwanda 65 · Tanzania 65

Employee bands: 11–50 (526) · 1–10 (306) · 51–200 (204) · 201–500 (136) · 1,001–5,000 (79) · 501–1,000 (64) · 10,000+ (44) · 5,001–10,000 (17) · Unknown (99)

Loading

from datasets import load_dataset

ds = load_dataset("mcsqstudio/africa-startup-directory", split="train")
print(len(ds))                 # 1475
print(ds.features)

Usage ideas

• Sector classification: predict operational_sector from company_description.
• Tier triage: predict aikya_tier from description + structured features.
• Market intelligence / gap analysis by region, sector and company age.

Missing data

• 1 record missing year_founded / company_age (COMP-000045, Agoro).
• 1 record missing hq_city (COMP-001390, Virtual City).
• 7 records missing operating_countries.
• aikya_tier/operational_sector fall back to Unknown where no label could be assigned (99 and 3 records respectively).

Files

File	Purpose
data.parquet	Canonical 1,475-record dataset (Parquet, nullable int32 years)
dataset_infos.json	Dataset schema & split metadata
push_to_hub.py	Script to upload this folder to the Hugging Face Hub

Ethical considerations

• Contains public company information only; no personal data or email/phone contacts.
• Some records are historical or flagged defunct (e.g. "Mobius Motors (Defunct)") — treat as a 2026 snapshot, not live company status.
• Verify licensing of any downstream redistribution that goes beyond academic/classification use.
• Company facts are as captured at collection time; verify before investment decisions.

Contact

MC Studio — Nairobi, Kenya. Sector/tier labeling follows the Aikya Legal (Okutta & Wairi Advocates) venture-support framework and the Sita Sector Program.