For the head of people analytics
You are asked for models while still reconciling headcount across three systems. This is the material on data contracts, metric definitions, attrition and adoption modelling, and how to make a number defensible.
How do we get from reporting to models people act on? A metric with two definitions is a metric nobody trusts. Definitions are cheaper to fix before the model than after it.
For the role
The same library, sorted by what you're actually responsible for.
Roles
One definition per metric
Write down owner, formula, source system and refresh for the ten metrics that reach the executive team.
A layer instead of exports
Move the ten metrics into one modelled layer with tests, so the same number appears in every report.
Forecasting on top
Attrition risk, demand and supply, adoption. Models built on the layer, not on spreadsheets.
28 poster
How far ahead can we realistically forecast headcount?
A direct answer on realistic headcount forecasting horizons, and why accuracy drops fast beyond a few quarters.
GuideWorkforce planning6 minScenario planning for workforce cost
How to build two or three workforce cost scenarios that give leadership a real choice, not a single forecast.
GuideWorkforce planning8 minWorkforce demand and supply forecasting, explained plainly
A plain-language guide to forecasting how much workforce you need and how much you will actually have.
GuideAdoption and change5 minA four-week pilot plan for an AI-in-HR tool
A week-by-week checklist for running an AI-in-HR pilot with a clear deliverable at each stage.
Why the pilot looked good and the rollout did not
Pilots often succeed because a small, motivated team gets close support and a manager who champions the change. Rollout removes both at the same time the audience grows, so the same tool meets less attention and more resistance. The fix is to remove support gradually across the rollout, not all at once after the pilot ends.
Human oversight, defined
A plain-language definition of human oversight as used in AI governance and the EU AI Act.
ChecklistGovernance and the EU AI Act6 minA DPIA checklist for people analytics projects
A working checklist for scoping a data protection impact assessment before a people analytics project goes live.
GuideGovernance and the EU AI Act7 minLogging and human oversight, built into HR agents from day one
A practical approach to designing logging and human oversight into HR agents rather than adding them later.
AnalysisSkills and workforce models6 minSkills data decays faster than most teams plan for
Why skills records go stale within months and how to set a realistic refresh cycle.
A skills inventory you can finish this quarter
Most skills inventories fail because they try to map every skill for every role at once. A narrower approach works better: pick one business question, map only the roles and skills that answer it, then expand. This delivers a usable inventory in weeks and creates momentum for the next round.
Which HR metrics should reach the board?
A short list of workforce metrics that belong in board packs, and why most dashboards do not qualify.
AnalysisPeople analytics and metrics5 minDashboards versus decisions
Why more dashboards rarely produce more decisions, and what actually closes that gap.
RegulationPeople analytics and metrics5 minDefining attrition so two teams agree
A concrete rule set for calculating attrition so HR and finance never present conflicting numbers again.
GuidePeople analytics and metrics6 minThe metric layer the people function runs on
How a shared metric layer turns scattered HR reporting into one language everyone can trust.
Readiness checklist before your first HR model
Before building an HR model, confirm the underlying data has stable worker identities, agreed field definitions, a known refresh cadence and documented gaps. Skipping this checklist does not remove the risk; it just moves the discovery of bad data from before the project to after it ships. Ten minutes of checking saves months of rework.
Worker identity across systems, not just one
Why a stable worker identifier matters more than any single system's own employee ID.
GuideHR data quality5 minData contracts between HRIS and analytics
A practical way to stop HRIS changes from silently breaking analytics and models downstream.
GuideHR data quality6 minWhat an HR data foundation actually is
The structural layer that must be true before any HR model or agent can be trusted.
ComparisonWorkforce planning6 minStrategic versus operational workforce planning
A comparison of strategic and operational workforce planning, and why organisations need both running at once.
Manager enablement: the step most AI-in-HR rollouts skip
Manager enablement means giving line managers a short, specific script for how their team's work changes, not just system access. Programmes that skip this step see technically live tools with no behaviour change, because managers keep approving the old way. Enablement content should be role-specific, delivered before go-live, and reinforced in the first four weekly one-to-ones.
Adoption rate is a model, not a feeling
How to define and track adoption rate as a measurable model instead of relying on impressions from a few power users.
RegulationGovernance and the EU AI Act8 minWhy employment counts as high-risk under the EU AI Act
What the high-risk classification for employment and worker management means in practice for HR teams.
GuideSkills and workforce models6 minThe task model: breaking roles into plannable units
How to break a role into tasks small enough to plan capacity, automation and hiring against.
ComparisonSkills and workforce models5 minSkills taxonomy versus job architecture
How a skills taxonomy and a job architecture differ, and why most organisations need both, in order.
What should never be automated in HR
Some HR decisions should stay with a human regardless of agent capability: termination, disciplinary outcomes, final pay decisions, and anything requiring judgement about a person's character or intent. Automating the analysis behind these is fine. Automating the outcome removes accountability from a decision that needs someone to own it.
Human-in-the-loop levels for HR agents
A four-level scale for how much human review an HR agent needs, from full sign-off to full autonomy.
CaseAgents in HR6 minInside an HR service desk agent, three months in
A synthetic case walking through an HR service desk agent in production, with the numbers behind its performance.
GuideAgents in HR6 minWhat is an HR agent, and where should it stop?
A plain definition of an HR agent, where it earns trust first, and the line it must not cross.
Want to see how this looks in your setup? The intro call takes 30 minutes, and we start with the People analytics view.
We read everything ourselves. No sales team in between.
From a first agent in production to the full people-function stack.
The full library, searchable and filterable.