Great Place To Work Hub: turnover and belonging, read together

Every large employer already holds the evidence. Who applied, who was advanced, who was promoted, who left and when — all of it sits in the human capital management system, because payroll and reporting require it. Separately, most of them run an engagement survey, which produces the other half: what people say about whether they can be themselves at work.

The two halves are almost never read together, and apart they are both survivable. A turnover number is a cost. A belonging score is a sentiment. Put them side by side and they stop being metrics and become a sentence about a specific group of people, which is a much harder thing to leave sitting in a dashboard nobody opens.

Great Place To Work Hub was the future vision for doing that inside UKG. I led the design.

Who is the readout actually for?

We started on that question rather than on the data model. The answer we designed against was not the executive who commissions the engagement survey. It was the manager of eleven people, who has never been given a reason to look.

The workshop sheet on the left is the whole method: identify the users, sketch against a theme for thirty minutes, prioritize five, then work the content areas. The sheet beside it is where the real argument happened — which practice areas an organization would trust enough to act on.

What we put on the screen

The hub opens on six scores, Belonging at 69 up to Pride at 92, with a Trust Index of 86. A score is where most products in this category stop. The design work was all of what comes after it.

Smart Insights reduced each area to one sentence with a group in it. Hispanic or Latinx employees aged 25 to 35 are 90 percent more likely to leave than other groups. Women aged 35 to 45 are 60 percent more likely to leave. Female employees in technical roles receive 80 percent of the bonus their male counterparts receive. Those sentences are the product. The charts underneath them are evidence for a claim that has already been made in words.

Every chart carries a plain-language reading. That is the accessibility requirement — a chart with color as its only channel excludes people before they have read a number — and it turned out to do a second job. A reading makes a finding arguable. A chart on its own is decoration that everyone in the room nods at.

One of those readings notes that more team members are choosing "prefer not to answer." That is a finding about the instrument, not the team. A workforce that has stopped answering the demographic question is saying something about how safe it feels to answer it. A tool that files that under missing data, rather than under signal, is measuring the wrong thing.

Why the hiring funnel was the hard one

Seven stages, 326 applications, 8 hires. The male share of the pool rises at every stage, from 60 percent at application to 80 percent at hire. Broken out by ethnicity instead, the white share grows through each stage, and white candidates are advanced about 15 percent more often than everyone else.

Nothing in that chart accuses anybody, and that is exactly why it works. Every one of those advancement decisions was made by a person doing their job, with a reason they could give you. None of them is the villain. The shape only exists once you stack them, which is precisely the kind of exclusion no individual reviewer is positioned to catch.

The design decision with teeth was where to put it. A funnel like that in a quarterly review is a slide. In the hub, sitting beside the reminder to schedule a one-to-one, it is in front of the person who runs the hiring process while they are running it.

What I cannot tell you

This was a future vision, and the word is load-bearing. It went through discovery interviews and focus groups, usability studies with HR executives from several organizations, and a full specification covering the hub and its Belonging, Hiring, Turnover and Promotions modals. I cannot point you at a shipped feature with a number attached.

The first write-up of this project listed its outcomes as "expected," which is a polite way of recording that nobody had measured anything yet. I would rather leave that plainly on the page than dress it up. Impact is not a fact until someone measures it, and a vision project is exactly where that distinction goes quietly missing.

Where this goes

The part I would take anywhere from this is the readout, not the dashboard.

Most organizations do not need more measurement of belonging — they are already measuring it, usually annually, usually well. What they lack is a moment when the measurement reaches the one person who can act on it. And a sentence about people, rather than a score out of a hundred.

That gap is ours. It is a design problem, in our category of software, about whose name appears in the sentence. If we build the readout so that the only honest way to read it is who, then the argument about what to do next has somewhere to start.