HR analytics: which metrics to track first

A metric with no decision attached is just a number

HR analytics usually starts from the wrong end: build the dashboard first, then look for something to do with it. It works better the other way round — start from the decisions the company makes regularly and keep only the metrics that move them. If nobody will act on a number, there is no need to compute it, however handsome it looks.

The second rule is comparability. A number on its own means nothing: whether 18% turnover is bad or good depends on the industry, the season and what it was last quarter. A metric becomes useful when it has a history and a breakdown.

Ten metrics worth starting with

They are chosen on one criterion: each can be computed from data the company already has, and for each it is clear who does what if it gets worse.

• Turnover by tenure — 30, 90 and 365 days. Three different causes, three different owners.
• Time to fill, from request to first day. Not to offer: more is lost between the offer and day one than people expect.
• Offer acceptance rate. A low one means you are competing on something other than you think.
• Cost per hire who actually started — counting managers' time, not just job-ad spend.
• Absenteeism: lost shifts as a share of planned, by site. A local spike almost always points at a manager.
• Overtime as a share of the wage bill. A rising share signals not busyness but a shortage on a specific shift.
• Labour cost as a share of site revenue. The one metric on this list a finance director will look at willingly.
• Share of roles filled internally. It shows whether people have a visible next step — and predicts turnover better than surveys do.
• Days to close the timesheet. A dull operational metric that explains half the pay disputes.
• A simple satisfaction score — one question a quarter. The value isn't the number itself but how it moves between teams.

How not to drown in dashboards

Hold to one rule: every metric has an owner and a threshold. The owner is a person, not a department. The threshold is the value at which they have to do something. A metric without a threshold becomes wallpaper: everyone sees it and nobody reacts.

And start with three metrics, not ten. Ten is where to arrive after a year, not where to sensibly begin: a team unused to deciding from data simply turns away from a large dashboard.