Making data meaningful: disaggregation and intersectionality

Why aggregated data obscures inequality

Headline statistics can be powerful tools for advocacy and communication. They can also be profoundly misleading. When data is reported only as averages or totals—across entire populations, regions, or demographic groups—the experiences of those at the margins become invisible.

Consider a health service that reports 85% patient satisfaction. This figure tells us nothing about whether satisfaction differs by ethnicity, disability status, language spoken, or neighbourhood. It tells us nothing about whether the 15% who are dissatisfied share common characteristics. It tells us nothing about whether the service is working for everyone or only for the majority.

Aggregation is not neutral. Decisions about how data is collected, categorised, and reported reflect assumptions about what matters and whose experiences count. When data systems are designed without attention to equity, they reproduce the blind spots of their designers.

The problem is compounded when aggregated data shapes decision-making. Policies designed around averages serve average populations. They systematically fail those whose needs, experiences, or circumstances differ from the norm. This is not an accident; it is a structural feature of how data is used.

Disaggregation as a justice and accountability tool

Disaggregation—breaking data down by relevant characteristics—is a fundamental tool for revealing what aggregation conceals. By examining outcomes across different groups, we can identify patterns of inequality that would otherwise remain hidden.

Disaggregated data answers different questions. Instead of asking "Is this service effective?", we ask "Is this service effective for everyone?" Instead of measuring average outcomes, we measure the distribution of outcomes and the gaps between groups.

This shift has implications for accountability. When organisations are required to report disaggregated data, they become visible for what they achieve—and fail to achieve—for different populations. Disparities that were previously hidden in averages become matters of public record.

Disaggregation also changes the conversation about improvement. Rather than generic initiatives to raise overall performance, it becomes possible to target interventions at specific gaps. Resources can be directed where they are most needed, rather than spread thinly across populations with unequal needs.

However, disaggregation is not sufficient on its own. Data revealing disparities must be accompanied by analysis that explains those disparities and action that addresses them. Without these, disaggregated data becomes documentation of inequality rather than a tool for change.

Intersectionality in practice, not theory

Intersectionality, as conceptualised by Kimberlé Crenshaw, recognises that people hold multiple identities simultaneously and that these identities interact to shape experience. A Black woman's experience cannot be understood simply by adding together the experiences of Black people and women; it is shaped by the unique intersection of these identities.

In data terms, this means that single-variable disaggregation—looking at outcomes by ethnicity or by gender—may still obscure important patterns. True intersectional analysis requires examining outcomes for people at the intersection of multiple characteristics.

Practical intersectional analysis faces genuine challenges. Sample sizes shrink as categories become more specific. Small numbers raise both statistical and ethical concerns: findings may be unreliable, and individuals may become identifiable. These challenges require careful methodological responses, not abandonment of the approach.

Strategies for managing small numbers include pooling data across time periods, combining similar categories where conceptually justified, using qualitative methods to supplement quantitative findings, and being transparent about limitations. The goal is to learn as much as possible about intersectional experiences while respecting statistical and ethical boundaries.

Intersectionality also requires attention to which characteristics are collected and how. Data systems that only record binary gender, that use broad ethnic categories, or that fail to capture disability status or socioeconomic position cannot support intersectional analysis. Data infrastructure is a precondition for intersectional insight.

Risks of misusing or over-simplifying disaggregated data

Disaggregated data, used carelessly, can cause harm. One risk is essentialism: treating observed differences between groups as fixed or inherent properties of those groups, rather than as products of social, economic, and historical conditions. Data showing poorer health outcomes for a particular ethnic group does not mean that ethnicity causes poor health; it reflects the accumulated effects of racism, discrimination, and structural disadvantage.

A related risk is deficit framing: presenting data about marginalised groups in ways that emphasise problems and pathologies while ignoring strengths, assets, and the structural causes of disadvantage. Disaggregated data should inform action to address inequities, not reinforce stereotypes about communities.

There are also risks of inappropriate comparison. Not all disparities are equally significant or actionable. Some reflect genuine differences in need; others reflect discrimination or poor service design. Interpreting disaggregated data requires contextual understanding, not just statistical analysis.

Finally, disaggregation without community involvement can feel extractive. Communities may be wary of data collection that documents their disadvantage without their input or benefit. Meaningful disaggregation requires partnership with the communities whose experiences are being measured.

Ethical considerations and community consent

Collecting disaggregated data requires trust. People must be willing to share information about their identity, which may be sensitive or contested. This trust must be earned through transparency about how data will be used, protected, and governed.

Consent should be informed and ongoing. This means explaining not just that data is being collected, but why disaggregation matters and how findings will be used to drive improvement. It means providing genuine choice about what information to share and respecting that choice.

Data governance should involve community voice. Decisions about what categories to use, how data is reported, and who has access should not be made solely by researchers or institutions. Community advisory structures, data sovereignty frameworks, and participatory governance models can all help ensure that disaggregated data serves community interests.

Protection of small groups is essential. When numbers are small, individuals may be identifiable even in aggregated reports. Suppression thresholds, data masking, and careful attention to indirect identifiers are necessary to prevent harm. The goal is to learn from patterns without exposing individuals.

How insight should travel back into action

Disaggregated data is only valuable if it leads to change. This requires clear pathways from insight to action: governance structures that receive and respond to disaggregated findings, accountability mechanisms that track progress on closing gaps, and resources allocated to address identified inequities.

Action should be proportionate to findings. Small disparities may warrant monitoring; large and persistent gaps require targeted intervention. The nature of intervention should be informed by analysis of causes, not just documentation of outcomes.

Communities whose data has been collected should see benefit from the findings. This might mean involvement in designing responses, access to findings before publication, or resources directed to community-led initiatives. Data extraction without reciprocity is not ethical research.

Progress should be measured and reported. Disaggregated baselines enable tracking of whether gaps are narrowing, stable, or widening. Regular reporting creates ongoing accountability and allows for course correction when interventions are not working.

What commissioners and analysts should demand from data

Commissioners of research, evaluation, and data analysis have significant power to shape how data is collected and reported. By specifying disaggregation requirements in contracts and specifications, they can ensure that equity is built into data practice from the start.

Specifications should require disaggregation by relevant characteristics, with clear rationale for which characteristics matter and why. They should require intersectional analysis where sample sizes permit. They should require transparent reporting of limitations and uncertainties.

Commissioners should also require interpretation, not just data. Reports should explain what disaggregated findings mean, what might be driving observed patterns, and what actions are indicated. Data without analysis is abdication of responsibility.

Analysts should push back against requests for aggregated-only reporting. They should proactively offer disaggregated analysis and explain its value. They should be transparent about what their data can and cannot reveal and advocate for improved data collection where gaps exist.

Both commissioners and analysts should invest in data infrastructure. High-quality disaggregated analysis requires high-quality data collection, which requires resources, training, and system design. Short-term cost savings on data collection create long-term blind spots in understanding.