Root cause analysis

When a number moves, find out why.

Ask why. IntRpret splits the change into parts that add up exactly to it and keeps the largest until they explain 80% of it, searches your graph's own signals for candidate causes, and tests each hypothesis as a measure — then goes a level deeper, up to four times.

Root cause · Regretted attrition, Sales3 tested
Moved · Q3 against Q211.8% → 14.6%
Where · 83% of the changeEnterprise team 52%Mid-market team 31%
Share paid below band midpointContributing
Internal transfers in the last 6 monthsCorrelated
Leave days taken per personUnrelated
Worded by the evidence — never called proven.Sample data
What it does

Root cause analysis, part by part.

  • 01

    Where it moved

    The change is split into parts that add up exactly to it, the largest kept until they explain 80% of it — and up to three places that carry it, at each level of the drill.

  • 02

    Searched, not listed

    Candidate causes come from your graph's own signals — HR events, features and record properties — never from a fixed list.

  • 03

    Hypotheses as measures

    Each possible cause is a condition found in the graph, shown as a measure, and classed a likely root cause, contributing, correlated or unrelated.

  • 04

    Worded with care

    Findings are worded by the strength of the evidence and never called proven; every step is kept, with the graph queries behind it.

How it works

Moved, Where, Test, Why, again.

  1. Step 1

    Moved

    The KPI's change, and the periods compared.

  2. Step 2

    Where

    The parts that carry 80% of it.

  3. Step 3

    Test

    Hypotheses found in the graph, each tested.

  4. Step 4

    Why, again

    It asks why of its own findings, up to four levels deeper.

The analysis, stage by stage

From a moved number to a cause

StageWhat it does
Moved
Measures the change and the periods it is between.
Dimensions
Chooses the ways to split it — department, grade, location, tenure.
Hotspots
Keeps the largest parts until they explain 80% of the change.
Discover
Searches HR events, features and record properties for candidate causes.
Classes
Tests each one and classes it a likely root cause, contributing, correlated or unrelated.
Confirm and keep
Checks the finding and keeps it with every step and its queries.
Ask it

Questions it answers, in your words.

  • “Why did headcount drop between January and February?”
  • “What is driving early attrition among new joiners?”
  • “Why did our cost per hire go up this quarter?”
Works with

On the same model and graph

Your next people decision, with the why attached.

Request a demo. We'll email you a two-minute walkthrough today, then set up a time that suits you.

We'll use your email to set up your demo. Remove it any time. Privacy Policy