Skip to content
Pawel Works

03 · Data

SQL / ETL, analytics backend

Analytics-ready models and pipelines: SQL Server, SSIS, Power Query. SAP, BW and Navision extracts, quality gates, and reconciliation before the dashboard exists.

Service · Transformations that survive the next source change

When this is the job

The report is slow because the model is a spreadsheet with extra steps. Grain is implied. Keys do not match. An ETL job fails quietly. Three systems (SAP, BW, Navision) dump files that nobody maps the same way twice. The dashboard is being asked to do work that belongs in the warehouse.

How it runs

Start at the grain. Entity structures, relationships, transform-and-load logic. File-based ingest from SAP, BW and Navision — content-detected, not filename-detected — with a quality gate before activation. SQL and Power Query that can be read. Tests at the edge cases. Document what the source will not give you. Optimize the query, not the visual.

  1. L1

    Ingest

    Content-addressed. A file is what it contains. Names are not a contract.

  2. L2

    Gate

    Keys, grain, nulls, duplicates. Nothing activates until it opens.

  3. L3

    Transform

    SQL you can read. Two grains, side by side.

  4. L4

    Serve

    A model a dashboard can trust. Tests at the edge.

What you leave with

A backend a dashboard can trust. Scalable transformations, reconciled outputs, and logic that does not live only in one person’s head.