What should we measure?

Problem

TDCNET had internal goals for who it attracts, hires and keeps, including hiring more women into a technical organisation. The raw numbers could not answer whether the company was on track. A share of new hires on its own says nothing until you know what share of the relevant graduates the company could have hired in the first place, and whether the people it hires also stay.

What I did

The method was the same for every goal, repeated across all of them:

1. Take one of the company's internal goals.
2. Find what the newest research and industry reports say drives it. A McKinsey report on what women value at work, for example, pointed to flexibility for parents.
3. Set up the numbers: where the company is today against where it should be, with a fair benchmark. For hiring, that meant comparing the share the company hired with the share of relevant graduates in public education statistics.
4. Check what the company offers on each driver, as a clue to why its numbers look the way they do.

Across all the goals that came to 27 internal and external datasets, followed over time and across business units, and checked against external market data before anything was presented to leadership.

How I decide what to measure

Most of the work is deciding what to trust. A number can look precise and still mislead: two datasets can define the same group differently, a small department can swing from year to year on two people, and a share of hires means little without the pool behind it. So every KPI got a benchmark, a definition that held across the datasets, and a check against an outside source.

The research did a different job. Comparing TDCNET's offer with what the reports said people value was not something I could test statistically, and I did not present it as proof. It was a clue to why the numbers skewed the way they did, which is often what leadership needs to decide where to act.

What came out

KPIs, trend analysis and recommendations for each target, presented directly to leadership. The point was not one headline number but a set of measures leadership could steer by: where the company was ahead of the benchmark, where it was behind, the likely reasons, and where to act first.

What I would add today

The same KPIs on a monthly refresh. Today I would pull the datasets automatically, run the same definition and benchmark checks every time, and put the results on one dashboard, so the analysis is spent on what the numbers mean rather than on rebuilding them.