Series
Getting your data in order
From loose exports to one version of the truth: master data, one source per field, pipelines, warehouse and migration, and why this decides whether AI works.
After this series you can:
- You point out where your data drifts apart and count how big the problem is, before you build.
- You record per piece of data which system is the source, who decides, and what is checked at entry.
- You pick what your data question needs: a pipeline, a warehouse or a migration, and require every step checked.
04 · Basic
Getting your data in order
From loose exports to one version of the truth: master data, one source per field, pipelines, warehouse and migration, and why this decides whether AI works.
What this series does not cover
- Which reporting or dashboard package to choose.
- What a warehouse, a migration or a clean-up costs.
- What the GDPR requires of your customer data. That is in the series 'AI and the rules'.
Terms
- data silo
- A system that keeps its data to itself, so questions across systems go unanswered.
- master data
- Data that rarely changes, such as customers and articles, used in every system.
- data quality count
- A count of duplicates, empty fields and wrong addresses, made before any cleaning.
- source system
- The one system where a piece of data is changed; the others only read it.
- entry check
- A rule that blocks or flags wrong data at the moment someone enters it.
- data pipeline
- A fixed, logged route that fetches, checks, converts and writes data for your reports.
- data warehouse
- One database that copies all your systems every night and keeps every version.
- trial migration
- A test run of the move on a copy, counted until every total matches.
