People Analytics

What people analytics teams actually do

Most of the work falls into four kinds of question, in rising order of difficulty:

1. What happened? Descriptive reporting: headcount, hires, exits, time to fill, absence, overtime. This is where HR metrics and dashboards live.
2. Why did it happen? Diagnostic analysis: which teams, managers, shifts, sources or pay levels are associated with higher turnover or lower performance.
3. What will happen? Predictive analytics: which new hires are at risk of leaving in the first 90 days, how many hires you will need next quarter, which requisitions will miss their target date.
4. What should we do? Testing and recommendations: pilots, A/B tests of process changes, and decision support that tells managers where to focus.

Most organizations spend most of their effort on the first level, and that is fine. Clean, trusted reporting is the foundation. Teams that jump to machine learning on messy data produce models nobody believes.

People analytics use cases that pay off

• Early attrition in hourly roles. A logistics company finds that new hires who are offered a shift different from the one they applied for leave at twice the rate in their first 60 days. Recruiting changes the offer process, and 60-day attrition drops.
• Source quality. A contact center compares hiring sources not just on cost per hire but on 90-day retention and QA scores. One job board looks cheap but its hires leave fast; employee referrals cost more upfront but stay.
• Assessment validation. TA checks whether pre-employment assessment scores actually predict performance and retention for each role, and whether pass rates differ by demographic group in a way that needs review.
• Manager effect. Analysis shows that a handful of team leaders account for a large share of regretted exits. That becomes a coaching priority, not a company-wide engagement program.
• Workforce planning inputs. Attrition curves by tenure and role feed the hiring forecast, so plans include realistic backfills.
• Pay equity reviews. Regression analysis of pay by role, level, tenure and performance identifies unexplained gaps to investigate with counsel.

Data sources

People analytics pulls from systems that rarely talk to each other: the HRIS (job, pay, tenure, manager, location), the applicant tracking system (sources, stages, time in stage, assessment results), payroll and time and attendance (hours, overtime, absence), performance and QA tools, learning systems, engagement surveys and exit interviews. In operations-heavy businesses, the richest data often sits in operational systems: contact center QA, productivity, scheduling and sales data.

The hardest part is usually joining these on a consistent employee ID and agreeing on definitions. If finance, HR and operations each define turnover differently, analysis turns into arguments about the numbers instead of what to do. A shared metric dictionary is the most valuable early deliverable a new team can produce.

People analytics vs HR reporting

HR reporting answers fixed questions on a schedule: monthly headcount, turnover, open requisitions. People analytics starts from a business question, like why are we losing night-shift agents in week six, and uses whatever data and methods it takes to answer it. Reporting is necessary but not sufficient. A team that only produces reports can tell leadership turnover went up; an analytics team can tell them where, why and what is likely to fix it. Custom dashboards sit in between, giving managers self-serve access to the most-used views.

Privacy, fairness and legal considerations

Workforce data is personal data, and some of it is sensitive. Limit access by role, aggregate results so individuals cannot be identified in small groups, and be clear with employees about what data is used and why.

Several laws apply directly. In the EU, the GDPR requires a lawful basis for processing employee data and gives employees rights over it. In California, the CCPA as amended by the CPRA gives employees and applicants privacy rights over their personal information. When analytics feeds decisions about hiring or promotion, US anti-discrimination law applies, and adverse impact against protected groups must be monitored. AI and automated decision tools face specific rules, such as New York City's Local Law 144, which requires bias audits and notices for automated employment decision tools used for hiring or promotion in the city. Rules vary by location and change often, so involve legal counsel early.

How to start a people analytics function

1. Pick two or three business questions leaders care about now, like early attrition or time to fill for one critical role.
2. Fix the data for those questions first. Clean definitions, consistent IDs, documented sources.
3. Deliver one useful answer within a quarter. A clear finding with a recommended action builds more support than a large platform project.
4. Hire for curiosity and business sense, not just statistics. The best analysts can explain a finding to an operations director in two minutes.
5. Measure the decisions changed, not the dashboards built.

For a practical walkthrough of the recruiting side, see this guide to data-driven recruitment.