What is the ‘Data Value Chain’?

Care is the step where one better understands the situation by empathising with the people and organisations affected. In other words, before you design a survey or registry, you take time to learn what people’s lived realities look like, how they experience exclusion, and where harm or invisibility might happen. Because people at the margins often bear the greatest risks when data is misused or misunderstood, ‘Care’ requires being transparent about potential ways data could be used and getting stakeholder buy-in on the process, and make sure they feel the future categories and outputs chosen will reflect their realities fairly. This is also a preliminary stage to reflect on what safeguards (privacy, do-no-harm, anti-discrimination) must be built into the tool and the overall governance process.

• You can carry out activities like Focus Group Discussions, community round-table meetings, key informant interviews (or other listening methods) to understand who is missing and why, before finalizing categories and questions.
• It is also useful to consult with policy analysts as they can help prioritize disaggregation by key attributes important to national development, thereby validating questions and methodology, which leads to tailored strategies for data collection.²⁷

Initiate is where we move from “we’ve heard the issues” to “we’ve set up the right group, values, and way of working to engage in this data collection process well.”

• The responsible institution can convene a small cross-sector working group (National Statistical Office, relevant ministry, CSOs, traditional leaders) and agree what “safe, inclusive data” means in this context.

Define is the step where institutions work together to identify and prioritise the specific issues to be addressed. It answers the questions, “What exactly are we trying to measure, for what decisions, and which intersections matter most in data collection?” to name a few.

• Given that we can’t disaggregate everything all at once, define is the moment where you choose, based on the previous steps: which identity axes are most decision-relevant (e.g., sex + age + disability + location), which groups are most likely to be missed, and what minimum categories you need so the data is usable without creating excessive burden or risk.

Define what “safe and useful publication” means before anything goes live. This is where you decide which disaggregations are decision-relevant (e.g., sex + age + location, or disability where feasible), and which breakdowns should not be released because they could increase disclosure risk or expose small or stigmatized groups. In other words, intersectionality here is also about making careful choices about where disaggregation becomes unsafe.

Develop outputs that people can actually use. Open data, in this framing, is not “open in name only.” To be usable and equitable, data should be shared in machine-readable, non-proprietary formats, and paired with comprehensive metadata so users can extract, analyze, and reuse it reliably. This is also where accessibility is even more important, which means considering the language(s) data is published in and whether it is accessible to the marginalised groups revealed in earlier steps.

• Release data disaggregated by at least two or three key identities (e.g., sex, age, urban/rural, disability type).

Review the publication with the people the data is about, and the people who will use it. Open Data Watch highlights that analysts should continue consulting community representatives so the analysis reflects lived experience and asks the right questions. This is also where you protect agency and reduce harm. Given that intersectionality shifts the balance of “data ownership” toward subjects, publications should respect their concerns and apply strong privacy protections through data security measures and anonymization methodologies.

Adopt safeguards where needed, so they aren’t optional or informal. This is the moment to formalize publication rules (e.g., anonymization methodology, suppression thresholds, access conditions for sensitive microdata) and provide responsible-use guidance so data isn’t easily misinterpreted or weaponised.

Initiate step: the responsible institution(s) will be building the bridge between producers and users (don’t assume uptake will happen). It’s best to bring together the people who turn data into decisions, such as policymakers, service providers, regulators, CSOs, community representatives, and more importantly, data intermediaries (journalists, civic tech actors, advocacy groups, policy units). These actors connect data to real audiences, translate technical outputs into usable narratives, and create feedback loops that encourage engagement and accountability.

Develop step: the responsible institution(s) could create the “translation layer” that makes intersectional patterns understandable. Given that uptake needs more than publishing and that it often needs products and spaces that help people use the data like briefs, dashboards, interpretation workshops, plain-language summaries, and community-facing formats. These resources would be considered data literacy and training.

Propose step: turns intersectional insights into options. This step calls for the translation of what the data shows into concrete decision options (policy change, targeting criteria, outreach strategy, service redesign, budget reallocation). This is where intersectionality matters most and where it’s best to ask what barriers only appear when identities overlap, and evaluate trade-offs so that solutions don’t help one group while leaving out (or harming) another.

Adopt step: makes inclusive use routine through incentives and governance. To prevent intersectionality from being optional, Open Data Watch and Data2X²⁴ point to practical levers such as: donors can require inclusive data use and fund skills; governments can embed it in guidelines, standards, and regulations; institutions can make engagement with affected groups a normal step in decision workflows. Adoption here is also about working together to choose the rules and incentives that make equitable uptake repeatable.

Deliver/Monitor step(s): requires the support implementation and track whether decisions reach the intersections. Once decisions are made, uptake continues through execution where the entity communicates what’s changing and monitors whether actions actually reach the intended groups with identities that compound to increase their state of marginalisation. If not, adjust (delivery channels, eligibility rules, accessibility, outreach partners).

Harvest/Understand step(s): treat uptake as a constant learning opportunity, not a one-off. Gather evidence on what users understood, what they misinterpreted, what they couldn’t access, and what communities experienced after actions were taken. Feed those learnings back into publication (clearer documentation, safer disaggregation) and collection (better categories, improved reach). This is how uptake becomes a reinforcing loop rather than a dead end.

Define → make success explicit across overlapping identities.
Define what “success” looks like in intersectional terms. Set indicators that track outcomes across overlapping identities (not just one dimension at a time), and specify which groups must see improved access or outcomes (e.g., rural women with disabilities without IDs; migrants in informal work; youth with psychosocial disabilities). This is where equity becomes measurable rather than assumed.

Monitor → track results and risks, not just reach.
Monitor implementation against the defined intersectional indicators, and also monitor harm. Using intersectional data responsibly means tracking both positive impacts and risks, especially privacy, exposure, stigma, and discriminatory outcomes. Clear ethical standards and transparent data governance help prevent harm, while ongoing monitoring allows quick adjustments when problems appear.

Harvest & Understand → pair numbers with lived experience, then adapt.
Harvest evidence from both quantitative monitoring and qualitative insight. Collect stories and other qualitative evidence from people at the “edges” of the system to catch gaps that indicators miss and to explain why a policy is working for some groups but not for others. Then make sense of what you’re learning and translate it into concrete changes—adjust targeting rules, redesign delivery channels, strengthen safeguards, or revise what data needs to be collected and published in the next cycle.

How this becomes a reinforcing loop (why it matters)
When people can see that intersectional data leads to real policy changes or improved services, trust grows, and so does demand for better data. Strong learning loops encourage continued participation, investment, and better practice, helping embed intersectionality into everyday decision-making over time.