What Word Captures a Decade in Data Visualization?

I’ve seen a shift away from New Year’s resolutions. Instead, people select a word to define a year (or a decade). Looking back over the past ten years in data viz, it feels like the 2010s were a decade of more—more tools, skills, data, and collaboration opportunities than ever before.

Over the past ten years working on data visualization in public health, I’ve been part of the increasing adoption and use of visualization to inform and delight. More charts and graphs, more dedicated education for public health practitioners, and more advanced analytics and visualization platforms, all fueled by more available data in the public health and clinical care areas.

Thinking forward to the next decade, the word I hope will define the 2020s in data viz—an intention, of sorts—is mindfulness. A shift away from the fervor around endless new tools and data sources towards careful consideration of what we visualize, who it represents, how it’s interpreted, and what impact those charts have.

How will these two periods be different for data visualization as a field of practice and study? What follows are my own observations from the past ten years, and ideas on what mindfulness means in data viz as we look forward.

2010s: A Decade of MORE

Increasingly Professional Data Visualization

Visualization has a long and storied history as a technique for communicating information. Throughout the last decade, I saw two transitions in public health as visualization evolved from a skill to a dedicated career path.

1 — People became experts (specifically) in data visualization.

Awareness of, and interest, in learning data visualization best practices expanded and was empowered by increasingly flexible, low-code (or no-code) tools. The emergence of dedicated health data viz professionals is complemented by a broader range of health professionals building charts and graphs in various tools. Often, data viz professionals have come from an adjacent discipline, like health IT, health information systems, monitoring and evaluation, research, and graphic design, bringing valuable insights from those domains.

2 — Organizations invested in their data visualization capabilities.

Some organizations have evolved to have dedicated data visualization or cross functional analytics teams, while others have worked to modernize how they visualize and use data by hiring dedicated data visualization consultants. The best kinds of consulting engagements leave the organization with more internal capacity for data viz design, and, perhaps more importantly, increased ability to use data for decision making.

This shift to data visualization as a profession matters. Names, roles, and defined teams help to codify and recognize the distinct blend of skills learned by studying and practicing the craft of data viz. Further, these new roles and teams reinforce the value these capabilities add to health organizations.

More Dedicated Analytics Training

Education programs have also evolved to meet the need for this focused skillset.

Dedicated data visualization classes are taught as part of some public health programs. Ongoing education initiatives (like the Population Health Exchange at my alma mater, BUSPH) feature workshops and courses on data visualization for public health. There are also virtual and in-person public workshops accessible to practitioners around the world.

More Low-Code Tools and Adoption of Visual Analytics

An explosion of low-code tools has enabled analysts to quickly create elegant charts and graphs with some embedded interactivity. Great examples include Datawrapper, Flourish, RAWgraphs, and Tableau Public. More complex chart types can be created by non-developers with tools like Sankeymatic, which provides in-browser editing panes for easy customization. These low-code tools promised to democratize the development of data visualizations and have made interactive design more accessible to analysts and designers.

The landscape of visual analytics tools for dashboard development expanded to include more robust interactivity features, mapping, and more. PowerBI joined Tableau in the Gartner Magic Quadrant, and competitors like Looker and Qlik continued to grow. Their popularity drove interest by other firms, and in 2019, Tableau and Looker were acquired by Salesforce and Google, respectively.

As the user base for these analytics tools has grown, the cost of the software has come down. Lower costs and dedicated programs for reducing licensing costs for small non-profits have made those analytic powerhouses more accessible for smaller organizations, translating to wider use (and impact) of dashboarding platforms. However, more dashboards hasn’t necessarily translated into wider data use.

More Influence from Design

When I was in my masters program (pre-dating this decade), we learned how to create basic graphs and charts to report on evaluation findings. There was a class on building dashboards in Excel for simple project monitoring that taught the functional basics of pivot table reports. None of our training focused on design.

More design influence has nudged health data viz developers toward designing for understanding, usability, and delight. Some of this evolution was empowered by improvements in widely used tools like Excel. The wide dissemination of best practices by data viz thought leaders in the health and social science spaces brought design best practices into the health sector (check out blogs like PolicyViz, Depict Data Studio, Evergreen Data, and Storytelling with Data). The impact of this information was accelerated further by increased collaboration with skilled design practitioners.

More Data

Open Government initiatives made the volume of deidentified health data and related social determinants information more widely accessible, unleashing many “citizen data scientists.” In the US, the move towards using EMRs (thanks to the Data Accessibility rules from the Affordable Care Act) meant that data previously stored in paper charts is now being managed in databases and warehouses.

More Collaboration, Conferences, and Knowledge Sharing

Local and global communities connected data visualization developers, making design best practices, code sharing, and support more accessible than ever. PUGs, TUGs, D3 groups, and local data viz meetups abound. There’s even a dedicated #projecthealthviz challenge every month that launched in 2018! In February 2019, the global Data Visualization Society launched, and hit the 10,000 member milestone in just 10 months, reaffirming the need for a broader, tool-agnostic community.

2020s: A Decade of Mindfulness

Following a decade of more people, more data, more tools, more communities, and more projects, we need to find ways to put this proliferation of opportunity to good use and find more meaningful ways to invest the time we spend visualizing data.

Let the 2020s be a decade of mindfulness in data visualization, where we evolve our practice to be more intentional in how we collaborate, design, and consider the wider implications of what data we make visual.

Mindfulness in How we Collaborate across Disciplines

The expansion of open data initiatives will continue to make more large data sets accessible. Cloud storage costs are declining, making it easier for organizations to store large, and growing, volumes of data. Continued use of EMRs expands the volume of individual patient data (still governed by HIPAA) being stored in machine readable formats. We’re certainly not facing a shortage of raw data.

Often the data we need to make meaningful health decisions is fragmented. A hospital’s systems of record for patient outcomes, finance, human resources, and other essential information live in different data silos. Even with advances in how data is captured and stored, we will still face the challenge of digitizing valuable historical data. Modernization of enterprise data systems will need to continue and evolve in order to make that data accessible for visualization.

Data viz developers shouldn’t operate in isolation (or act as data unicorns). We need to collaborate across data disciplines (data engineering, data science, data visualization, and data analysis) and adjacent areas (design, software development, and health IT) to achieve our big goals and allow data to be used to drive organizational decision making, policy, increased quality of care, and improved health outcomes.

Mindfulness in What Data we Visualize

We need to amplify and expand our thinking about the people represented in the data, those who are missing from the data, and how the population represented would feel looking at what we designed.

Books like Invisible Women (2019) and Data Feminism (2020) shine a light on the biases in large data sets, and potentially dangerous unintended consequences. Integration of data science and advanced analytics with data visualization platforms can help make the output of algorithms and complex analyses more accessible, but we must be mindful in how we visualize that information. Research has shown that visualizations imply certainty and a basis in fact, which isn’t always the case (see How Charts Lie).

Mindfulness around Who is Reading our Visualizations

Over the last decade, the influence of design in our practice has pushed visualization developers to start to think first about their audiences, rather than jumping into development work.

Understanding who is reading our charts and considering their data literacy should be part of every visualization design process. We imply certainty when we plot data in bars, lines, and other encodings.

Being mindful of who is consuming our visualizations pushes us to think beyond enterprise dashboards and scorecards of “big” data. An expanded range of bespoke and custom visualization tools will empower patients to visualize and explain their own health care journeys.

Firms like Pictal Health are innovating with how to visualize small sample and individual health data. As the democratization of data visualization continues (see low code tool explosion in 2019), I’m hopeful that more patients will be able to visualize their own data and use those charts to facilitate conversations with caregivers across our fragmented health system. The most meaningful charts are often the ones where you see yourself in the data.

The dialogue around visualization research and visualizing uncertainty will continue well into the next decade, and perhaps we’ll see wider adoption of relevant techniques that help to communicate uncertainty without confusing the reader and address challenges with small sample size visualizations. We have, perhaps, over-adopted the Tufte adage about eliminating chart junk. Sometimes, the “junk” is what helps our audience make sense of what we’ve created or provides visual clues about the uncertainty around the data points.

Mindfulness around the Ethical Implications of our Design Choices

Last but not least, data ethics must inform our visualization design practices. The onus is on us as developers to question the motives and morals behind how our visualizations get used. Just because we can visualize a data set, should we?