People Analytics for Non-Analysts: A Starter Guide

People analytics has acquired an intimidating reputation — machine learning, attrition-prediction models, dashboards with regression lines. That image stops most HR teams from starting. It shouldn't: the majority of the value in people analytics comes from counting things accurately, cutting them sensibly, and asking one more question than usual. A capable HR generalist with a spreadsheet can do all of it.

Start with questions, not data

Bad analytics begins with "what can we report?" Good analytics begins with a decision someone needs to make:

• Should we raise engineering bands, or is our attrition there about something else?
• Is our hiring process slow everywhere, or only for certain roles?
• Which teams should worry us in the next six months?

Write down the five questions your leadership actually argues about. That is your analytics agenda.

The toolkit is smaller than you fear

Four moves cover most real-world people analysis:

• Count accurately. Headcount, exits, offers, joiners — with clean definitions, consistently applied. Most "analytics problems" are actually definition problems.
• Segment. Almost every insight lives in a cut: by tenure, manager, function, performance tier, compensation position. An average hides; a segment reveals. (Our piece on attrition diagnostics is essentially this move applied to turnover.)
• Trend. A single number means little; the same number across eight quarters tells a story. Keep definitions stable so the line is real.
• Compare. Against your own history first, market benchmarks second. Internal comparisons — team A versus team B — are usually more actionable than industry percentiles.

Three analyses worth doing this quarter

• Regretted attrition by manager and tenure band. One pivot table. It will surface your two or three real retention problems.
• Hiring funnel conversion by stage and role family. Where do candidates fall out — screening, interviews, offer? The bottleneck is rarely where people assume.
• Compensation position versus performance. Plot (even roughly) where your high performers sit against band midpoints. High performers paid below market is your resignation-risk list.

Habits that make data trustworthy

• One source of truth per metric, with a named owner.
• Definitions documented in plain language — what counts as an exit, a hire, a requisition.
• A monthly thirty-minute review where someone asks "what changed, and why?" Insight comes from the conversation, not the dashboard.

The traps for the enthusiastic beginner

• Correlation theatre. "Teams that did the offsite have higher engagement" — or happier teams chose to do offsites. Hold conclusions loosely; triangulate with interviews.
• Tiny-sample confidence. A team of six with two exits is a 33% attrition rate and zero statistical meaning. Note small bases explicitly.
• Privacy carelessness. Small segment cuts can identify individuals. Apply minimum group sizes everywhere, and treat people data with the same care as financial data.
• Dashboards without decisions. If a metric has not influenced a decision in two quarters, retire it from the review.

Where this leads

Teams that build these habits earn something larger than better reports: credibility in business conversations. When HR speaks in segments, trends, and costed options, it stops being a support function and starts shaping strategy — the substance of HR's seat at the table.

If you want help setting up the first useful dashboard and the review rhythm around it, get in touch. The build typically takes a few weeks, not a transformation programme.