Humanitarian Dashboards

As an Emergency Data Analyst for UN Women, the first layer of this work was review, not code: Gaza emergency site-sweep XLSForm materials, humanitarian stakeholder correspondence, and existing public Power BI outputs were assessed against data-sharing requirements to map which gender-responsive indicators and priority request fields were safe to expose publicly, and where a schema field itself carried protection-sensitive risk that meant it couldn't be published at all, however useful it might have been analytically. That review produced the reliability caveats that shape everything downstream — a dashboard for crisis-response analysis has to be honest about what its inputs can and can't support, not just visually polished.

The dashboards themselves are built with Python, JavaScript, Chart.js and MapLibre for the mapping layer, with automated validation scripts running against every data refresh rather than trusting a source feed to stay clean. Multilingual UI support, comparison tables across regions and time periods, interactive map layers, and methodology/source panels next to every chart are treated as first-class requirements, not polish added at the end — a humanitarian audience reading a dashboard needs to see where a number came from as readily as the number itself.

Underneath the UI sit reproducible pipelines that integrate nine public evidence layers — WFP/HDX price data, the IPC food-insecurity classification, World Development Indicators, FAOSTAT, UNHCR and OCHA FTS displacement and financial-tracking data, ILOSTAT labour statistics, IDMC/IOM displacement figures, and Global Findex financial-inclusion data — into a common structure for prioritising support across crisis-affected populations. On top of that data, a documented Shock Exposure Index defines how component indicators are scored and weighted, how missing data is treated rather than silently dropped, before/after food-price screening around a shock event, and gender and labour vulnerability indicators, together with explicit interpretation rules that keep the index read as prioritisation evidence rather than a causal claim about why a shock happened.

Because this work supports partners who don't all use the same tools, delivery isn't a single dashboard link: outputs are packaged as handover-ready Excel, CSV, JSON, Word, PDF, PowerPoint and static-web artifacts, with source inventories and data dictionaries so a colleague picking up the workbook later knows exactly where each column came from. Governance sits alongside the outputs rather than after them — source-of-truth documentation, output manifests, data-retention rules, dry-run audit scripts, and workspace health reports are part of the same delivery, aimed at keeping the work reproducible and accountable as it's handed off across a humanitarian programme.