Election Watch
Election Watch was a pair of election analytics dashboards designed to help media members interpret the 2020 U.S. presidential race. Hawkfish’s models incorporated expected mail-in voting to show how an apparent election-night lead could change as more ballots were counted.
I led product design on the Election Watch project, translating complex election data into intuitive visualizations, reusable patterns, wireframes, and interactive prototypes. I worked with the development team through handoff and implementation, adapting designs when changes were needed under tight deadlines.
This case study focuses on how I designed the interaction models for two stages of the election, and used user feedback to make the context behind the data visible.
The challenge
Hawkfish’s data pointed to a potential “Red Mirage”: Donald Trump would appear to lead on election night before the unprecedented volume of mail-in ballots cast during the pandemic was counted, shifting the result toward Joe Biden. With the count expected to continue for days or potentially weeks, an early lead could be mistaken for a final outcome, influencing public sentiment and response.
Before the election, media members needed to explore scenarios and understand how mail-in voting could affect the race. Then, once counting began, their attention would shift to current results, unresolved states, and counties where the remaining votes could matter most.
With that in mind, the Election Watch experience needed to support both tasks while keeping the distinction between early results and projected outcomes clear.
Product direction
Competitive analysis showed that existing election dashboards offered curated scenarios and controls for adjusting partisan support at the state level, but did not incorporate mail-in voting as an interactive parameter, or provide county-level exploration.
Hawkfish’s data models offered an opportunity to address those gaps. The product direction called for a pre-election dashboard that the media could use to focus on various what-if scenarios, followed by a post-election dashboard focused on ongoing results and curated analysis.
I would use different approaches for the two tools:
The pre-election dashboard needed to make our curated scenarios easy to compare, clearly identify the projected winner in each, and show how mail-in votes could change the outcome through before-and-after views.
The post-election interface needed to establish the current state of the race, then support easy investigation of consequential states and counties.
Pre-election Design strategy
Given the intended audience of media members, we took a desktop-first approach, which would provide enough space to present detailed data clearly.
For the pre-election dashboard, I prioritized navigability of our curated what-if scenarios and showing users the projected winner of the scenario at a glance, with granular controls tucked away for advanced users.
Interacting with the scenario selector to produce a small moment of delight.
To show the state-by-state breakdown of the selected scenario, I explored a single map with a toggle that switched between them, highlighting the states that changed. This approach gave the map a large, dedicated area while keeping the layout compact.
An early design with a “Red Mirage” toggle, highlighting states that would flip after mail-in votes were counted.
Usability research
Usability testing validated the scenario-switching feature: users could easily navigate high-profile scenarios and immediately understand their projected results.
However, the map showing the selected scenario’s state-by-state breakdown received mixed feedback. We expected users to switch between views showing results before and after mail-in ballots were counted, but testing revealed that the interaction was unintuitive.
User feedback also revealed a broader risk: screenshots of the default view could circulate without the projected outcome after mail-in ballots were counted, stripping away essential context and potentially reinforcing a misleading impression of the race.
Responding to the feedback, I revised the design to show the two maps side by side. The modeled election-night picture and projected final outcome were visible together, with no interaction required to reveal the comparison. For users who wanted to adjust the models for deeper exploration and more nuanced scenario customization, the interface also included advanced controls behind a floating action button that revealed the controls alongside the selected scenario’s data sources. This kept detailed configuration accessible without competing with the primary comparison.
The updated layout used more space, but addressed the context missing from the default view, making screenshots more accurate and shareable.
Side by side map views showing before and after votes by mail (VBM) are counted.
Post-election design strategy
The post-election dashboard shifted the emphasis from exploring possible scenarios to understanding the ongoing count.
The default view prioritized the current state of the race and key national data, showing notifications of newly called states and refresh countdowns to keep the status of the information visible.
Notification examples on the post-election dashboard.
Curated daily narratives directed attention toward unresolved states and counties that could affect the result, adding historical context from previous elections. Users could select a state to examine county-level voting data or choose a narrative to follow Hawkfish’s interpretation of salient data points.
A curated narrative showing the remaining states to be called, depicted visually by electoral weight.
I worked with the team to refine a compact visualization, internally called the “Mondrian,” that summarized Trump and Biden’s voting data alongside Hillary Clinton’s 2016 results. Applied at national, state, and county levels, it provided a consistent way to read the same types of information at different scales.
The Mondrian, as dubbed internally (left), was deceptively simple, but conveyed key pieces of data at a glance.
A configurable county table supported deeper analysis, with more than a dozen data categories that users could show or hide. Details such as relative population, partisanship, and demographics complemented the curated narratives, giving users both a guided view and access to the underlying data.
One of multiple views of the county table, configurable to hide/show over a dozen categories of data.
The resulting experience
The final designs established two distinct ways to work with election data: comparing modeled scenarios before the election and examining evolving results during the count. The interaction model kept the main view focused while providing access to more detailed controls and county-level information.
Testing changed a foundational part of the pre-election dashboard design. Showing the maps together made the projected outcome visible alongside the modeled early lead, addressing the risk that the default view could be misunderstood. In the post-election experience, curated narratives, consistent visual summaries, and detailed tables supported different depths of analysis.
The dashboards’ effect on media coverage was difficult to measure. Internally, the project brought Hawkfish’s data, technology, product, and design teams into close coordination on a shared product for the first time.
The resulting design made context part of the experience itself, from the side-by-side scenario maps to the national, state, and county views used to follow the count.
The final versions of the pre- and post-election dashboards.