Predictive HR Analytics
One book, placed in its field — what Dr Martin Edwards's book gets right, where it goes further than the rest, and what the rest of the field adds.
Mike West, Editor · Leading With Data · peopleanalyst.com/guides/profile-predictive_hr_analytics_mastering_hr_metric
Arrow keys or swipe to move · N for notes · F for full screen
Predictive HR Analytics
One book, placed in its field — what Dr Martin Edwards's book gets right, where it goes further than the rest, and what the rest of the field adds.
Mike West, Editor · Leading With Data · peopleanalyst.com/guides/profile-predictive_hr_analytics_mastering_hr_metric
The problem
Employee / Team Performance erodes when it is left to instinct instead of method.
You were taught the moves piecemeal, never the whole model.
Predictive HR Analytics · Mike West, Editor · Leading With Data
The model
- 01HR Intervention / Programme
- 02Selection and Onboarding Practices
- 03Job and Team Characteristics
- 04Demographic and Diversity Composition
- 05Employee Engagement
- 06Perceived Organizational Support and Justice
- 07Job Strain / Stress
- 08Job Satisfaction
Predictive HR Analytics · Mike West, Editor · Leading With Data
Step 1 of 5 · HR Intervention / Programme
Design the evaluation of an intervention before the intervention, not after the results disappoint you.
Misconception
That rolling out a training or wellbeing programme and seeing engagement rise afterwards proves the programme worked.
Reality
Post-launch improvement is usually confounded by seasonality, concurrent changes, and the fact that motivated employees self-select into voluntary programmes; without a comparison group you are measuring who enrolled, not what the programme did.
Predictive HR Analytics · Mike West, Editor · Leading With Data
Step 2 of 5 · Selection and Onboarding Practices
The quality of a selection prediction is capped by the quality of the outcome label it was trained on.
Misconception
That assessment-centre scores and aptitude tests are objective predictors of performance because they produce numbers.
Reality
These instruments predict whatever the historical performance labels they were validated against actually captured — often manager favouritism or tenure — so a high-scoring predictor can simply be reproducing who succeeded under old conditions, not who will succeed under new ones.
Predictive HR Analytics · Mike West, Editor · Leading With Data
Step 3 of 5 · Job and Team Characteristics
The demand–control interaction, not demand alone, predicts who burns out.
Misconception
That job demands and job control are just background variables to control for, when the real action is in individual attitudes.
Reality
Job control is a moderator, not a nuisance: the same level of demand produces high strain in low-autonomy roles and manageable pressure in high-autonomy ones, so context changes the sign and size of relationships in your model.
Predictive HR Analytics · Mike West, Editor · Leading With Data
Step 4 of 5 · Demographic and Diversity Composition
Excluding protected attributes from features does not prevent proxy-driven bias; you must measure impact on the labels you excluded.
Misconception
Practitioners believe that dropping gender, ethnicity and age from the feature set makes a model fair and legally defensible.
Reality
Proxy variables — zip code, alma mater, prior tenure gaps, commute distance — reconstruct protected status even when it is excluded, so blindness masks bias rather than removing it. You must retain the demographic labels precisely so you can measure disparate impact you would otherwise never see.
Predictive HR Analytics · Mike West, Editor · Leading With Data
Step 5 of 5 · Employee Engagement
Engagement correlates with performance and turnover; do not sell it internally as a proven cause of either.
Misconception
Practitioners treat engagement as a direct cause of performance, assuming a rising engagement score will pull productivity up with it.
Reality
Engagement and performance mostly co-vary because a shared cause — good jobs, fair treatment, capable managers — drives both; the relationship in the model is a correlation, not a proven lever, and reverse causation (success breeds engagement) is real.
Predictive HR Analytics · Mike West, Editor · Leading With Data
A framework to use
The Predictive HR Analytics Project Framework
This framework is part of the full guide. See the guide →
Predictive HR Analytics · Mike West, Editor · Leading With Data
The cheat sheet
- HR Intervention / ProgrammeDesign the evaluation of an intervention before the intervention, not after the results disappoint you.
- Selection and Onboarding PracticesThe quality of a selection prediction is capped by the quality of the outcome label it was trained on.
- Job and Team CharacteristicsThe demand–control interaction, not demand alone, predicts who burns out.
- Demographic and Diversity CompositionExcluding protected attributes from features does not prevent proxy-driven bias; you must measure impact on the labels you excluded.
- Employee EngagementEngagement correlates with performance and turnover; do not sell it internally as a proven cause of either.
- Perceived Organizational Support and JusticeProcedural fairness usually beats distributive generosity as a driver of retention — invest in transparent process, not just richer outcomes.
- Job Strain / StressStrain predicts turnover and performance most sharply when demand is high and control is low — model those two dimensions separately.
- Job SatisfactionThe behavioural-intent component of satisfaction predicts turnover far better than the affective feeling component.
Predictive HR Analytics · Mike West, Editor · Leading With Data
Sources
- Predictive HR Analytics — Dr Martin Edwards
Predictive HR Analytics · Mike West, Editor · Leading With Data