Data Design Guidelines

We will not share personally identifying information about donors

This includes providing "anonymous" data with sufficient fidelity such that others could make inferences about donors. Donors will be grouped into meaningful categories, and we will only display summaries across these categories. This requirement limits the granularity of our categorization (e.g., geographic data at the census block level). This also precludes the use of conditional categorization (i.e., further categorizing donors who all share a secondary trait), since it could be used to identify people.

We will only use meaningful donor categories

We should pay close attention to user input and find out what people want to know about the donors supporting candidates. However, for us to include a donor categorization in our interface, it must be meaningful.

Meaningful categories are those that help differentiate candidates and help users identify those that are aligned with their interests. For example, if the distribution of donation amounts is very different between candidates running for the same City Council seat (Candidate A receives mostly small donations, Candidate B receives mostly large donations), then donation amount would be a meaningful way to categorize donors. However, if--for example--the geographic distribution of donors is identical within sets of competing candidates, including geographic distribution will only clutter the interface.

We strive to use visualizations that are not misleading

Users should be able to accurately infer the information used in any visualization. This does not require a prescribed set of visualization rules (i.e., all bar charts must have a y axis starting at zero); rather, this goal can be met through a combination of following best practices and performing ad hoc user tests.