In my opinion, there are two aspects that make data warehousing of HR data, and therefore, people analytics, remarkably complex and elusive.
1. HRIS systems store transactions much differently than what is needed for reporting
Transactional systems have databases that are built to store and retrieve a transaction in the most efficient way possible - not to aggregate data for reporting. If someone was adding a bunch of transactions to an HRIS (e.g., bringing on summer hires), the technical result will be the addition of numerous rows to the database - something that an HRIS is set up to do.
On the other hand, a data warehouse in a people analytics solution stores data much differently than any transactional system. It stores data centered around the employee, position, or requisition, versus storing the data around a system transaction like a hire, promotion, or pay raise.
The logic needed to translate a single dimension transaction into a people-based record (that identifies key characteristics about an employee and his/her position) is really quite complex.
2. You must also construct this people-based record by bringing together different data from different sources within the context of point-in-time
As discussed in point 1, because data warehouses store differently than a transactional system, when we bring data together in an HR data warehouse, the technical activity shouldn’t be viewed as "adding transactions" as much as it is “adding more context” to an employee, position or requisition.
Time is one of these contextual dimensions - and creates a number of challenges when designing, building, and optimizing the performance of an HR data warehouse.
So when bringing in other data sources into a data warehouse for HR reporting and people analytics, we are transforming them to fit within the context of the data that is already there - which is this people-centered view.
Let’s take performance data, for example, which is often from another system or possibly even a spreadsheet. For this data, we are not just taking a file and inserting records (as your transactional systems are designed to do). We are assessing the time period for which this rating is valid (starting with an Effective Date) and adding in that characteristic, within that specified time period, to an existing employee.
This means that there’s lots of under-the-surface magic which must occur so that when you go back in time to analyze the impact that accelerated pay increases have on performance and productivity, you’re data is structured in a way that will make that analysis easy and speed-of-thought.
So when it comes to true people analytics, data preparation, storage, and data management, you need two components:
- Very capable tools focused on data management and discovery, not transactional processing
- People skilled and experienced in the dark-art of HR data warehousing
This need has given rise to the HR Tech segment called the Multi-Source People Analytics Platform - learn more about this and the people analytics technology landscape through RedThread Research or by reaching out to us here at PeopleInsight.