Payroll open-data dashboard for Antigua Guatemala
An interactive open-data dashboard that lets the citizens of Antigua Guatemala see and understand how their municipal payroll is spent — automated integration, AI-assisted cleaning, and accessible visualizations.
- ~95% reduction in time for citizens and journalists to interpret payroll
- Public-facing dashboard for civic transparency
- Foundation for predictive payroll-trend analytics
Context
Open government data only works if the citizen on the other end can actually read it. The Municipality of Antigua Guatemala publishes its payroll — that part is real transparency. But the format is closer to "obligation met" than "insight delivered." Without analytics chops or hours to spare, a resident or a local journalist cannot meaningfully ask: who in this department earns what, how has it trended, where are the outliers?
This dashboard exists to close that gap. The same payroll data, rendered as a public-facing exploration surface that a non-technical citizen can navigate in minutes.
Role
Data Analyst and Product Lead. The pipeline that feeds this dashboard is documented in the companion data engineering case study; this work is the analysis layer and the public-facing visualization layer on top of it.
What we built
Three layered pieces:
- Automated data integration. A streamlined pipeline keeps the dashboard fed from the cleaned payroll dataset as soon as a new period is ingested. The information citizens see is current, not stale.
- AI-assisted classification and cleaning. Job titles, department names, and role mappings in raw municipal data are inconsistent — different spellings, different categorizations across periods. AI-driven classification resolves these into a stable taxonomy so trends over time are real, not artifacts of a renamed cost center.
- Interactive dashboard visualizations. Bar charts for cost-center breakdown, trend lines for period-over-period change, drill-downs by department and role. Every view is built to be explorable by someone who has never opened a BI tool.
Product decisions worth writing down
Design for the journalist, not the analyst. The primary user is a citizen with an investigative question, not a finance team with a saved query. Every interaction defaults to "can a curious non-technical person find a story in five minutes," and the visualizations were tuned against that bar.
Taxonomy stability over time matters more than precision in a single period. A 2% misclassification in one period is forgivable. The same role labeled three different ways across three periods is fatal — citizens lose the ability to track change. The AI cleaning layer optimizes for stability first.
Explainability of the underlying data is part of the design. Every chart links back to the source rows in the cleaned dataset and through to the original municipal PDFs. A citizen who wants to verify a number can.
Outcome
The dashboard reduced the time citizens and journalists need to interpret municipal payroll data by approximately 95%. The work strengthens the public's ability to engage with how their tax money is spent and sits inside a broader Open Government Data thread that runs through the rest of my research. Next on the roadmap: extending into budgets and expenditures, and layering predictive analytics so the public can see where payroll trends are heading, not just where they've been.