African Last-Mile Logistics & Delivery Dataset
5M+ GPS-tracked delivery records and driver logs from e-commerce operators across Nigeria, Kenya, and South Africa — including geocoding quality scores, traffic-adjusted ETAs, and failed-delivery reason codes for training route-optimisation, ETA prediction, and address-normalisation AI.
This is a synthetic dataset generated from high-quality expert-labelled seed data. All records are algorithmically derived — statistical distributions, inter-field correlations, and annotation characteristics faithfully replicate real-world patterns from the source data, while ensuring no real individual, organisation, or transaction can be identified or reconstructed.
The African Last-Mile Logistics & Delivery Dataset captures 5M+ individual delivery attempt records collected from e-commerce and quick-commerce operators across Lagos, Nairobi, and Johannesburg / Cape Town. Each record tracks the full lifecycle of a delivery order from dispatch to final outcome — recording depot departure time, GPS waypoints at key events (dispatch, arrival at delivery zone, delivery attempt, return to depot), traffic-adjusted ETA at dispatch, actual delivery time, and the outcome code (delivered, failed, rescheduled, returned).
A distinctive feature of the dataset is its geocoding quality layer. African addressing is highly informal — many delivery zones rely on landmark descriptions, area codes, and driver local knowledge rather than structured street addresses. Each record includes a geocoding confidence score (0–1), an address type classification (STRUCTURED, LANDMARK, AREA_CODE, FREE_TEXT), and a normalised coordinate pair derived from a best-effort geocoder pipeline. Failed deliveries carry a structured reason code from a 12-class taxonomy (customer absent, address not found, access denied, security zone, etc.) enabling targeted model development for specific failure modes.
The dataset supports route optimisation under realistic African urban constraints — traffic congestion, road quality variation, fuel cost sensitivity, and address ambiguity. It is structured as a time-stamped event log compatible with common logistics ML frameworks and geospatial libraries (GeoPandas, H3, OSRM). A companion road-network graph (OpenStreetMap-derived) is provided for routing model development.
Key Use Cases
Dataset Highlights
Geographic Coverage
Dataset Schema
Each record represents one delivery attempt. Fields cover order identity, timing, GPS tracking, address quality, and outcome classification.
| Field Name | Type | Description | Nullable | Example |
|---|---|---|---|---|
| delivery_id | STRING | Unique delivery attempt identifier | No | DLV-NGA-LG-20230914-0041823 |
| order_id | STRING | Parent order identifier (multiple attempts may share this) | No | ORD-NGA-0918234 |
| attempt_number | INTEGER | Delivery attempt sequence number for this order (1 = first attempt) | No | 1 |
| country_code | STRING | ISO 3166-1 alpha-2 country code | No | NG |
| city | STRING | City of delivery | No | Lagos |
| dispatch_timestamp | DATETIME | ISO 8601 dispatch datetime from depot | No | 2023-09-14T08:32:00Z |
| eta_at_dispatch_min | INTEGER | Traffic-adjusted ETA in minutes at time of dispatch | Yes | 47 |
| actual_delivery_min | INTEGER | Actual time from dispatch to delivery outcome in minutes (null if unresolved) | Yes | 63 |
| address_type | ENUM | Address format: STRUCTURED, LANDMARK, AREA_CODE, FREE_TEXT | No | LANDMARK |
| geocode_confidence | FLOAT | Geocoding confidence score (0–1, higher = more precise location) | No | 0.71 |
| dest_lat | FLOAT | Destination latitude (WGS 84, best-effort geocoded) | Yes | 6.4281 |
| dest_lng | FLOAT | Destination longitude (WGS 84, best-effort geocoded) | Yes | 3.4219 |
| outcome | ENUM | Delivery outcome: DELIVERED, FAILED, RESCHEDULED, RETURNED | No | DELIVERED |
| failure_reason | STRING | Failure reason code from 12-class taxonomy (null if delivered) | Yes | null |
| driver_id | STRING | Anonymised driver identifier | No | DRV-NGA-0841 |
| vehicle_type | ENUM | Delivery vehicle: MOTORCYCLE, CAR, VAN, BICYCLE | No | MOTORCYCLE |
Sample Records
Four representative delivery records spanning countries, address types, outcomes, and vehicle types.
Build with Data that reflects Africa
Request access to our full catalog of licensed human-validated African datasets or request custom data tailored to your project.