IntermediateHRPeopleOps
Broken HR advice #9: 'Be data-driven about people decisions'
People data has the lowest signal-to-noise of any business data. 'Data-informed' is honest. 'Data-driven' is a hiding place.
7 min read
On this page▾
60-Second Summary
- Most HR datasets have sample sizes too small for the inferences leaders try to draw — a 200-person company has 8 promotion decisions a year, not a statistically meaningful population.
- The 'data-driven' framing is often used to defer judgment calls and to absorb accountability for a decision made for political reasons.
- Use the term 'data-informed' — it preserves the role of judgment and assigns the accountability where it belongs.
If you have 200 employees, you do not have a dataset. You have anecdotes arranged in a spreadsheet.
The sample-size problem
n=8
Promotion decisions per year, 200-person company
Below threshold for statistical inference
n=18
Annual attrition events at 200 ppl
Too small to segment meaningfully
12 months
Time before a hiring decision shows true quality signal
Most 'hiring data' is pre-signal
Data-informed vs data-driven
Two postures
Data-driven (overclaim)
- The data showed X, so we did X.
- Decision authorship hidden behind dashboard.
- Hard to overturn even when the data is wrong.
Data-informed (honest)
- Here's the data, here's the judgment we layered on it, here's who decided.
- Decision authorship explicit.
- Easy to revisit when context changes.
Further reading
Written by Pawan Joshi.Sources cited inline.
First published 10 Feb 2026See site changelog →
Related
Read next
All playbooksHR Analytics
Broken HR advice #5: 'Run an annual engagement survey'
Annual cadence is the wrong cadence for the only signal that actually predicts attrition. Here's the cadence that does.
7 minIntermediate
HR Analytics
Real attrition cohort math: what 42 hires over 36 months told us
Three years of monthly cohorts, plotted, with the actual exit reasons and the math behind the second-year wall.
13 minAdvanced
HR Analytics
A/B testing HR interventions: experimentation for people programs
How to apply experimental design to HR — when randomisation is possible, when to use quasi-experimental methods (difference-in-differences, regression…
17 minAdvanced