← Back to insights
Strategy

Numbers show who is in the room. They do not show who belongs.

By Angel Samuel Perez · · 4 min read

Rows of seats in an empty auditorium

Your diversity data can improve while the people it describes feel less included than ever. Representation is easy to measure. Belonging is what matters.

Representation statistics remain the default measure of DEI progress in most organisations: headcounts by demographic category, tracked over time, presented as evidence of change. This is not without justification. But the research is consistent on one point: representation, taken alone, is a poor proxy for inclusion, and can actively obscure the problems it is meant to reveal.

A useful illustration comes from outside the corporate world. A 2025 study published in Frontiers in Psychology examined diversity metrics in Major League Baseball, an industry that reports high representation of international Latino players and has, on that basis, been treated as a DEI success story. The researchers found that despite this high diversity, some players and coaches had less ability to advance than others, with international Latino players less likely to reach the highest levels of the sport than US-born players, and international Latino coaches more likely to remain stuck at lower levels. A supplementary survey found that non-White players reported a lower sense of inclusion than their White counterparts, despite the sport's strong representation numbers. The headline statistic and the lived experience of the workforce were, in other words, telling two different stories.

This gap is not incidental. It reflects a conceptual distinction that researchers in organisational behaviour have been trying to formalise for some time. A systematic review of measures used to assess organisational inclusion, published in the Journal of Business Diversity, identified five recurring themes across the academic literature: equal treatment from peers and supervisors, inclusion promoted at the organisational level, empowerment, recognition of individual uniqueness, and belongingness. None of these can be inferred from a headcount. Representation tells you whether a group is present. It says nothing about whether its members are treated equally, able to contribute, or made to feel that their particular way of working and being is welcome rather than merely tolerated.

Belonging in particular resists quantification in any direct way. Research on inclusion in technical and organisational communities describes it as a theoretical concept that is hard to observe directly, though it can be approximated through survey questions grouped into a broader latent construct. That word, latent, is doing real work. Belonging is not a number that exists in the world waiting to be counted, the way headcount is. It has to be inferred indirectly, carefully, and usually through methods that are qualitative in nature: interviews, structured listening exercises, open-ended survey items that let people describe their experience in their own terms rather than selecting from a predetermined scale.

The implication for practice is not that quantitative metrics should be abandoned. They remain useful for exactly the things they are good at: identifying where a pipeline is leaking, tracking whether a specific intervention is moving a specific number, and creating the kind of accountability that a qualitative account alone struggles to provide. The error is treating them as sufficient. Reviews of the field consistently frame the task as bridging qualitative and quantitative research, rather than choosing between them, and note that oversimplified interpretations of complex identities are among the more persistent failure modes when organisations rely on one type of data alone.

The baseball case is a useful corrective precisely because it is so legible. A sport with strong representation numbers, held up as a model, turns out on closer inspection to have real disparities in advancement and a measurable gap in how included different groups feel. The number was not wrong. It was simply answering a narrower question than the one the organisation thought it was asking. Anyone using representation data to evaluate inclusion should ask what that data cannot tell them, and treat that gap, not the topline figure, as the place where the real work begins.

References