Skip to content
Empley

The metric layer the people function runs on

6 min readPublished 2026-09-07Last reviewed 2026-09-07
Short answer

A metric layer is the shared set of calculations, sitting between raw HR data and every dashboard or report, that defines exactly how each number is produced. It stops finance, HR and managers from each computing attrition or headcount a different way. Every people analytics capability, from dashboards to predictive models, depends on this layer being correct and stable.

How to do it, in short
  1. 01Inventory current metricsCollect every existing definition of the core metrics in use today.
  2. 02Reconcile differencesPick one formula per metric and document why alternatives were rejected.
  3. 03Centralise the calculationCompute each metric once, upstream, and have every dashboard read from that single source.

People analytics is often judged not on sophistication but on whether two people looking at the same metric get the same number. That trust is built, or lost, in the metric layer.

What sits in the metric layer

  • The formula for each core metric, such as attrition, span of control and time to fill
  • The population each metric applies to, including exclusions like interns or contractors
  • The time window and rounding rules used consistently everywhere

Why dashboards alone are not enough

A dashboard displays a number. It rarely documents the formula behind it in a way a second team can replicate. When two dashboards use different logic, both can be technically correct and still disagree.

We had three attrition numbers in one deck. Nobody could say which one was right.
Head of People Analytics, mid-size technology company

Building the layer

A metric layer is unglamorous work, but it is what allows every later analytics or AI capability to be trusted at face value.

In this topic

Underlag

  • Our assessment

    Conflicting attrition or headcount figures between HR and finance are a frequently cited source of executive distrust in people data.

Common questions

Is a metric layer a piece of software?
It can be implemented as a semantic layer in BI tooling, but the essential part is the agreed definitions, not the tool.
How many metrics need to be in it initially?
Most organisations start with five to ten core metrics that appear in leadership reporting.
Next step

Want to see what this looks like in your organisation? An intro call takes 30 minutes and we reply within 24 hours.

Book an intro call

We prefill the request for you.

See all services

Three entry points and five modules, from a first process to an operating model.

See the agent roster

The agents that staff each step, with a clear mandate and review point.