Know in advance instead of explaining afterwards
A forecast is only useful when it feeds a decision: a supplier order, shifts, a budget. We first check there is enough history, then build the model.
01Builder
Sketch your data architecture
Pick sources and goals — the diagram appears on the right straight away: sources, pipeline, warehouse, BI and outputs.
Preliminary architectureNodes: 10
- ERP · 1C → ETL · ELT
- Billing → ETL · ELT
- Website · events → ETL · ELT
- ETL · ELT → Data quality checks
- Data quality checks → Data Warehouse
- Data Warehouse → Data marts, Forecast models
- Data marts → BI layer · metrics
- BI layer · metrics → Dashboards
- Dashboards
- Forecast models
SOURCES ERP · 1C · Billing · Website · events PIPELINE 3 connectors · Scheduled / incremental loads · Deduplication · validation · reference data STORAGE Data Warehouse (PostgreSQL or ClickHouse by volume) · Data marts per department OUTPUTS Forecast models · Dashboards
This is a preliminary outline. The architecture may change once we look at your sources.
02Forecasting
What we forecast
Sales and demand with seasonality and promotions, operational workload, churn risk, deviations from normal.
04Contact
Let’s look at your data
Five questions and you get a preliminary data architecture and a first step. Or just write to us.
Preliminary data architecture
Step 1 of 6