How can data visualizations represent the complex mosaic of public opinion? Headlines focus on the big polling numbers, but what about the real numbers on the ground? How do opinions vary from neighborhood to neighborhood and county to county? True Views (trueviews.org) is a mapping platform that answers these questions in a beautiful and engaging way.

Using complex statistical models, political scientists at Harvard Law developed a fine-grained analysis of Americans' beliefs across 32 different policy areas. With years of data and over one million responses to analyze, a ZIP code level model was developed, allowing the public and policymakers to visualize how opinions vary across the map.

To enable the exploration of this massive dataset, we (greeninfo.org) built a lightweight mapping application that is optimized for performance and accessibility. The vector tile data format (PMTiles) was crucial for allowing users to rapidly explore data of this size. These tiles allow the storage of both geographic and tabular data, so the browser only has to download a few small files for the map to render. Each shape contained the numbers for each policy question, allowing the cartography to change instantly based on user input. To enable this speedy design, we used a robust data pipeline (bash, Python) to pre-process and upload these datasets to the cloud, avoiding the need for slow databases and APIs.

But it’s not just about speed: from day one, we built TrueViews to be fully keyboard navigable, ensuring that people of all abilities can explore the data for themselves. Where possible, WCAG best practices were followed, and data was always accessible in multiple ways—both in a table and on the map.

As we developed the application, we explored the data and discovered surprising common ground on topics portrayed as divisive in news headlines. If we uncovered value and meaning while developing the tool, we knew that this tool could be a powerful catalyst for change.

In this talk, I will share our process from start to finish, starting with design exploration, data pipeline development and tools, and implementation tips for speed and accessibility.