1Bronze
Raw data
Untransformed ingestion using MinIO and Apache Iceberg.
PHASAO innovation centre
ANStat Data Lab
Innovation and technological sovereignty at the heart of the National Statistical System.
Purpose
Our mission
Created within the Economic Analysis Unit (CAE), the Data Lab is ANStat’s operational arm for innovation. Its mission is to modernise national statistical production by using mobile, satellite and web data, with three goals: accuracy, responsiveness and cost efficiency.
Technology foundation
A sovereign Data Lakehouse architecture
A unified pipeline turns raw data into ready-to-use indicators with data auditing, traceability and version control.
2Silver
Clean data
Typed, deduplicated and enriched data using Apache Spark and Airflow.
3Gold
Ready for use
Business indicators, federated SQL queries and dashboards using Trino and Superset.
A fully containerised Kubernetes deployment keeps data in Côte d'Ivoire, auditable and secure.
PHASAO innovation centre
Available equipment
- Sovereign MinIO storage and Apache Iceberg tables
- Distributed Apache Spark processing and Airflow orchestration
- Trino federated queries and Superset dashboards
- Secure containerised infrastructure running on Kubernetes
PHASAO innovation centre
Services provided
- Continuous statistical data flows complementing field surveys
- Predictive modelling using AI, NLP and remote sensing
- Automated collection, processing and visualisation
- Indicators for decision-makers, researchers and citizens
Innovation portfolio
Five execution pillars
Five distinct data flows feed a unified architecture and complement traditional surveys with continuous, cost-efficient monitoring.
Satellite poverty mapping
- Measure poverty between major national surveys conducted three to five years apart.
- VIIRS, Sentinel-2, Sentinel-5P, VGG16 and Transfer Learning.
2021 estimate: 37.86% versus 37.50% measured in the field; R² ≈ 0.40.
LMIS 2.0 — Labour market radar
- Detect emerging skill and recruitment trends in Côte d'Ivoire in near real time.
- Daily scraping of 5 platforms, Spark, Airflow and NLP occupation coding.
Fully automated pipeline and an interactive dashboard tracking emerging skills.
COICOP — E-commerce prices
- Reduce price collection delays for the Harmonised Index of Consumer Prices.
- Scraping Jumia, Glovo, Batinet Prestige and Aliwax, followed by French-language classification using CamemBERT.
More than 92% accuracy in automated product classification under COICOP.
DTM — Mobility data
- Map internal migration and territorial dynamics using mobile phone data.
- Apache Spark, SQL and a privacy-preserving architecture with Orange CI, MTN CI, Moov Africa, ARTCI and the World Bank.
Interactive mobility maps and regional dashboards are being prepared.
ANStat-IA — Conversational assistant
- Make thousands of pages of official statistical reports published each year easier to use.
- RAG, ChromaDB, a specialised LLM, Python, LangChain and Django.
Instant answers in French supported by 100% verifiable official sources.
Tangible effects
Results achieved
The five data flows converge into the National Statistical System, turning raw data into validated and certified intelligence.
More targeted and responsive public policies for Government
More accurate monitoring of development indicators for partners
Greater transparency and accessibility for citizens and researchers
High-frequency statistical production complementing field surveys
Next step
Development outlook
Short term: deploy mobile phone metadata indicators in production and launch the ANStat-IA beta. Medium term: extend Data Lakehouse governance across the NSS and automate Python/R pipelines at scale. Long term: expand South-South cooperation and position ANStat as the ECOWAS statistical innovation hub.
Ecosystem