Building a People Analytics function from zero: a 24-month playbook

60-Second Summary

People Analytics has a four-stage maturity model: Operational (counting things), Advanced (segmentation and dashboards), Strategic (predictive + experimental), Transformative (embedded in business decisions). Most companies stall at stage 2 for years.
First hire = a People Analyst who is 60% data scientist, 40% HR business partner. Not a pure data scientist (won't understand the business questions), not a pure HRBP (can't build the models). The hybrid is rare and worth paying for.
Foundations to build in year 1: clean HRIS data, defined people taxonomy (level, role family, location), a single source of truth for headcount, and a quarterly people-data audit. Without these, every analysis fights the data.
Four use cases that buy credibility in year 1: workforce planning model, manager-effectiveness diagnostic, retention deep-dive with predictive component, and the company's first true A/B test of an HR program.
Common failure modes: hiring a data scientist who can't talk to HRBPs; building dashboards no one reads; modelling without an intervention plan; chasing AI before fixing data foundations.

Almost every company above 500 employees says they want People Analytics. Most build a function that produces dashboards no one reads and reports no one acts on. This is the 24-month playbook for building one that actually changes business decisions.

The 4-stage maturity model

Where you are determines what to do next

1
Stage 1 — Operational
Headcount reports, basic attrition, ad-hoc requests. Reactive. 'How many people work in Sales?' Most pre-500-employee companies live here.
2
Stage 2 — Advanced reporting
Dashboards, segmentation, KPI scorecards. Quarterly people review. 'Attrition by tenure and function, with trendline.' Most 500–5,000-employee companies stall here for years.
3
Stage 3 — Strategic analytics
Predictive models, causal analysis, A/B testing, workforce planning models. People analyst embedded with HRBPs. 'These 50 employees are at highest risk of leaving in Q3.'
4
Stage 4 — Transformative
People analytics is part of how the business runs — embedded in product decisions, revenue forecasting, M&A diligence. CHRO has analytics partner the way CFO has FP&A. Rare even in F100.

Bersin's maturity index data: ~70% of companies sit at Stage 1–2. Stage 3 correlates with 30% higher revenue per employee. Stage 4 is rare and disproportionately Big Tech.

First hires and team shape

Stage	Team size	Composition
First 6 months	1 person	Senior People Analyst (the hybrid — see below)
6–18 months	2–3 people	+ Data Engineer (HRIS/data pipelines), + part-time HRBP partner
18–36 months	3–5 people	+ Senior data scientist, + reporting analyst, + visualization specialist
3+ years	6–12 people	Sub-teams: analytics products, research/experimentation, ops & data engineering

The unicorn first hire

Hire one person who is genuinely 60% analytical (SQL, Python or R, regression, basic ML) and 40% HR business partner (knows the difference between regretted and unregretted attrition, understands comp bands, can sit with a VP of Sales). Pure data scientists from product orgs typically fail; pure HRBPs can't build the models. Pay 20–30% premium for the hybrid — it pays back 5x.

The minimum tech stack

HRIS as source of truth: BambooHR / Rippling / Workday / SAP SuccessFactors / Hibob. The choice matters less than the discipline of treating it as the single source.
Data warehouse: Snowflake / BigQuery / Redshift. People data lives next to finance and sales data — never in a separate HR silo.
Pipeline: Fivetran or Airbyte to pipe HRIS, ATS, performance, engagement, payroll into the warehouse.
Modelling layer: dbt for transformations and a canonical people-data model.
Analytics: Tableau / Looker / Mode / Hex / Power BI. Whichever the rest of the company already uses — never adopt a separate tool for HR.
Python/R notebook environment (Jupyter, Hex, Deepnote) for ad-hoc analysis.
Survey tools: Culture Amp / Glint / Lattice / custom — exported into the warehouse, never trapped in the tool.

Data foundations in year 1

People taxonomy: define level (L1–L8 or equivalent), role family (~30 across the company), location, function. Inconsistent taxonomies destroy 80% of analyses.
Single source of truth: declare HRIS the master. Resolve all disagreements (finance headcount vs HR headcount vs IT-account-count) within 90 days.
Historical data: load at least 24 months of monthly snapshots into the warehouse. Without history, no trend analysis, no models.
Engagement integration: pipe survey results into the warehouse linked at employee level (with proper access controls). Surveys trapped in tools are nearly useless analytically.
Quarterly data audit: 1-day exercise checking duplicates, missing fields, taxonomy drift. Skipping this for two quarters is enough to break models.

The 80/20 of year 1

80% of year-1 work is data engineering and taxonomy. 20% is analysis. Teams that flip this ratio (hire data scientists to build models on broken data) waste their first 18 months.

Year-1 use cases that buy credibility

The four that consistently work

1
Workforce planning model
Quantitative model connecting hiring plan, attrition forecast, recruiter capacity, and ramp curves to a quarterly capacity forecast. Used by finance and CHRO. (See workforce-planning-math article.)
2
Manager-effectiveness diagnostic
Composite score per manager: team engagement + team retention + team performance distribution + 360 + span. Used in calibration and manager development. Most political; build last in year 1.
3
Retention deep-dive with predictive layer
Cohort attrition analysis + logistic regression model. Identify the top 10–20% retention risks and the drivers. Used in targeted retention conversations.
4
The first true A/B test
Pick one HR program (manager training, new onboarding track, comp comms variant). Randomise. Measure. Publish results — including if it didn't work. Establishes experimentation muscle and proves People Analytics isn't just dashboards.

Year 2: predictive and experimental

Operationalise the attrition model: monthly scoring, top-decile flagging, manager toolkit (with ethics guardrails — see predictive-attrition article).
Run 2–4 HR experiments per year on real programs. Build the muscle of testing, not just measuring.
Introduce ONA (likely survey-based first) for one strategic question (M&A integration, RTO audit, succession risk).
Embed an analyst with each major HRBP team. Stage 3 only happens when analysts sit with HRBPs and pre-empt questions.
Publish a quarterly People Analytics review — methodology, findings, decisions taken. Builds external credibility and prevents the function being seen as a black box.

Common failure modes

Failure mode	What it looks like	Fix
Data scientist with no business sense	Builds models nobody uses; can't translate findings	Hire the hybrid; partner data scientists with HRBPs
Dashboards no one reads	Beautiful tableau workbooks, 4 views/month	Embed metrics in existing reviews (QBRs, ops reviews) instead of standalone dashboards
Modelling without intervention	Attrition model with 0.82 AUC and zero retention conversations	Pair every model with a defined intervention path before building
Chasing AI before fixing data	GenAI HR copilot on broken taxonomy	Year 1 is plumbing; year 2 is analytics; AI is year 3+ for most companies
Reporting to wrong leader	Buried in HRIS / Total Rewards team	Should report to CHRO or to a CoE lead with a direct line to CHRO
No data ethics framework	Surprise revelation, employee revolt	Publish use cases and safeguards from day 1; review quarterly

FAQ

Frequently asked questions

How much should we budget?

Year 1: USD 250–500k loaded cost for first hire + tooling. Year 2: USD 800k–1.5M for a 3-person team + warehouse + analytics tools. Below this, you're building Stage 1 capability and calling it Stage 3.

Should People Analytics own engagement surveys?

Usually yes — they own the measurement instrument, the analysis, and the reporting. The action plans stay with HRBPs. Splitting these between teams creates the 'measurement vs action' dysfunction that kills survey programs.

When should we hire a CPO/Head of People Analytics?

Around 2,500–5,000 employees, or earlier if data is a strategic priority. Below that, a Senior Manager / Director-level People Analytics lead reporting to the CHRO is the right scope.