What an HR data foundation actually is
An HR data foundation is the set of shared definitions, identifiers and update rules that let every system agree on who an employee is and what happened to them. Without it, headcount, attrition and cost figures diverge by team. Models and agents inherit whatever inconsistency sits underneath them, so fixing the foundation is cheaper than fixing every downstream report.
- 01InventoryList every system that stores employee data and what it claims to be authoritative for.
- 02ReconcilePick one definition per core field and document the exceptions.
- 03EnforceRoute updates through the system of record instead of editing copies directly.
Every people analytics project eventually stalls on the same question: which system is right when two reports disagree? That question is a symptom of a missing data foundation, not a dashboard problem.
What counts as a foundation
A data foundation is not a warehouse. It is agreement: one worker identifier across systems, one definition per core metric, and a documented rule for when records update.
- A stable worker ID that survives job changes and system migrations
- Shared definitions for start date, termination date and employment type
- A single source of truth per data domain, with the rest treated as copies
Why models and agents need it first
An AI agent answering a manager's question about attrition will quietly pick whichever definition its source system uses. If that definition disagrees with finance's, the agent becomes a source of confusion rather than trust.
Signs the foundation is missing
- Headcount differs between HRIS and finance by more than a rounding error
- Two people teams report different attrition rates for the same period
- Nobody can name the system of record for a given field without checking with a colleague
Where to start
This groundwork is unglamorous, but it is what determines whether a later AI investment compounds or has to be redone.
In this topic
Data contracts between HRIS and analytics
A practical way to stop HRIS changes from silently breaking analytics and models downstream.
AnalysisWorker identity across systems, not just one
Why a stable worker identifier matters more than any single system's own employee ID.
ChecklistReadiness checklist before your first HR model
Nine checks to run before commissioning any predictive or generative HR model.
Underlag
- Our assessment
Most HR analytics delays trace back to unresolved identity or definition conflicts rather than tooling gaps.
Common questions
- Is a data foundation the same as a data warehouse?
- No. A warehouse stores data; a foundation defines what that data means and who owns it.
- How long does building a foundation take?
- Small organisations can reconcile core fields in weeks; complex, multi-system estates often take a quarter.