Data-driven HR: the metrics that matter

Most HR teams already collect numbers. The harder part is asking a question the numbers can answer, and then doing something with the answer. That gap is the whole distance between reporting and analytics.

Four levels, and where most teams stop

Descriptive analytics tells you what happened: headcount, turnover rates, training hours. Diagnostic analytics tells you why, which usually means slicing the same data differently, turnover by manager, or the correlation between engagement and performance.

Predictive analytics estimates what will happen, through flight risk models or projections of where you will run short of talent. Prescriptive analytics suggests what to do about it, from the compensation mix to the intervention most likely to keep someone who is at risk of leaving.

Most teams live on the first level and call it analytics. The jump worth making is the second one, because that is where a number starts pointing at a decision.

The metrics worth tracking

For workforce planning: total turnover under 15%, voluntary turnover under 10%, regrettable turnover under 5%, plus the time it takes to fill an open role.

For performance and development: goal achievement above 75%, internal promotion above 70%, and training that gets applied on the job above 60%.

For engagement and culture: eNPS above 30, engagement above 4.0 out of 5, recognition happening more than once a month per person.

For diversity, equity and inclusion: a pay equity gap under 5%, and promotion rates that hold up across groups.

Where this usually goes wrong

Five failure modes account for most of it. Vanity metrics that look good in a board deck and change nothing. Correlation read as causation. Averages that hide the two groups they are made of. Data collected and hoarded, never shared with the managers who could act on it. And the most common one of all: a genuine insight that nobody turns into a decision.

Sign up for a free TalentRewards trial.