Guide: Data Quality

Goal

Deliver fast without building on low-quality data.

SharePoint / file shares at large scale

When connecting large SharePoint estates:

• do not ingest everything at once: select relevant sites/scopes first
• reduce duplicates/outdated content: less noise, better retrieval
• use clear metadata/naming conventions: better findability

Pragmatic sequence:

1. define top use cases
2. map only relevant data scopes
3. expand step by step after validation

SAP / ERP with many tables

For very large table landscapes (e.g. SAP):

• do not start with full coverage
• curate tables by use case
• assign business owners per data domain

Recommendation:

• start with a small core set
• validate answer quality
• expand table scope in controlled increments

Minimum standards for structured data

• stable keys/IDs available
• consistent date fields
• null/empty handling is understood
• field semantics are documented
• clear update cadence (e.g. hourly/daily)

Minimum standards for document data

• clear titles/file names
• current versions over shadow copies
• avoid legacy archives in first scope
• consistent folder/metadata structure

Go/No-Go checklist before pilot

• Is first scope clearly bounded?
• Are data owners assigned?
• Are 1-2 high-value use cases explicitly defined?
• Is it clear which data is intentionally excluded from phase 1?

If these points are clear, pilot speed and stability improve significantly.