Assignment 3: Visualization Critique & Redesign

1.0 Overview

In January 2022, the New York Times published an opinion article by Dr. Jeffrey Shaman featuring the spiral visualization shown in Figure 1. It caused some controversy. In this assignment we read the visualization, critique its design choices, and develop a redesign through sketches and a final polished chart.

2.0 Reading and critique

Before critiquing the chart, it is useful to restate the context and what the article is trying to argue. The article’s core claims are: (1) Omicron is more contagious than previous variants, (2) U.S. case numbers tend to be worse in winter than in other seasons, (3) this winter wave will produce record case counts, and (4) Omicron appears less severe, so deaths and hospitalizations should be proportionally lower. The feature visualization should support some of these claims, while also meeting an implicit requirement: (0) it must be visually striking and suitable for a square homepage thumbnail (Figure 2).

2.1 Insights into the data

On first examination we can make some inferences as to what is being displayed.

• There are clear peaks around July 2020, December 2020 to February 2021, August 2021 to October 2021, and especially January 2022. These appear as segments where the red band is much thicker than surrounding months.
• Low-case periods are visible early in the pandemic (January to March 2020), around March 2021, and around June 2021, where the band becomes thin.
• A recurring seasonal structure is visible: lower case load in spring to early summer, growth through late summer to autumn, and strong growth in winter.
• The latest portion of the spiral (late 2021 into early 2022) visually dominates, supporting the narrative that the current wave is unusually large.

2.2 Design critique

This is a unique graph which is relatively effective at bringing across several of the article’s points. (1) The visualization also strongly communicates that case load is higher “now” than earlier phases of the pandemic: the most recent segment is markedly thicker than prior turns of the spiral, giving an immediate sense of escalation. (2) Aligning months across successive years makes seasonal structure legible: the reader can directly compare “December-to-February” from one year to the next and infer likely near-future behavior. For instance, the elevated winter period in December 2020 to February 2021 is echoed by the rise in December 2021 to January 2022 (where the plot ends), which supports an expectation that February 2022 could be worse still. That seasonal comparability would be harder to achieve in designs that do not explicitly align months. (0) Finally, as a feature graphic it succeeds on attention: it is visually distinctive, uses a bold, constrained palette, and the continuously expanding spiral reads as an emotive metaphor for the repeating cycle of waves—here, it’s bad, it’s better, it’s back.

Despite these strengths, several design choices reduce legibility and limit what the chart can support. The most consequential issue is geometric: the spiral appears “off” and visually unstable, which makes careful year-to-year and season-to-season comparisons harder and also makes the chart feel unintentionally sloppy. As discussed by EagerEyes and reproduced in Figure 3, the baseline deviates from the Archimedean spiral many readers implicitly expect, producing an off-center, “wonky” appearance that competes with the goal of readable comparison. It is also not immediately obvious where to begin reading, and the chart relies on subtle directional markers (arrows) to communicate the traversal direction. This represents an additional cognitive step, especially since circular conventions are common for hours but less common for a calendar year. The increasing radius introduces a perceptual bias: later months occupy more visual space, so 2021 can feel “bigger” than 2020 even before considering the data, and comparisons become easiest only at the same angular position (e.g., April 2020 to April 2021) while becoming much harder across different months (e.g., April 2021 vs July 2021). This is further exacerbated by symmetric thickness above and below the baseline, which encourages area judgments rather than length judgments and makes magnitude estimation less precise. In terms of the article’s argument, the chart primarily reinforces (1) higher case load and (2) seasonal variation, but it does not directly address (4) reduced severity of the current wave and only indirectly supports (3) projection of high future cases, projected record cases. Fortunately, we now have an opportunity to improve.

3.0 Visualization sketches

Sketch 1: Stacked line chart

To begin, I ‘unwrapped’ the spiral graph from the article and placed it onto a more traditional line chart. This makes it easier to compare month-on-month and year-on-year statistics. My horizontal axis is continuous with month-based divisions. I would use the first day of each month as the tick location; however (as with the original spiral), this makes resolving week-scale divisions complicated since there is nothing that explicitly states this is what is happening. The vertical axis is continuous and linear, and I elected to write out the full number (i.e., 1,000,000) to help convey the scale of the data. I used color to highlight the 2022 spike and progressively less striking colors (here green and blue; in the final version likely shades of grey) to show the 2021 and 2020 plots. I used a simple line for 2020 and 2021 but appended an arrow to 2022, connoting the idea of continuing time and potentially explosive future growth in case load.

There are several drawbacks to this visualization. Since we’re in the northern hemisphere, winter straddles the uncomfortable 2020/2021 and 2021/2022 divide, making the winter peak harder to observe and making the chart appear more like a bathtub plot with a flat middle and rising edges. Again, as with the spiral graph, we have no way to communicate the severity of the disease.

In reference to the main purpose of our article and chart:

0. Visually interesting: Partially successful. The multiple colored lines are engaging and hint that compelling inferences can be drawn. The sharp spike in 2022 is especially striking. Modifying the horizontal scale allows us to achieve the desired 1:1 resolution for use on the NYT website. However, this is still ultimately a standard line chart and is unlikely to generate many clicks.
1. More contagious: Very successful. We see a clear spike in 2022.
2. Worse in winter: Partially successful. We can see correlations between years, but without added emphasis the reader must make this inference independently.
3. Potential for explosive growth: Successful. The arrow in 2022 helps to connote explosive growth.
4. Less serious strain: Unsuccessful. There is no information about the severity of the strain on this plot.

This is a good start, but we can improve. In my next sketch I want to make strain seriousness and winter peaks much more explicit, while limiting the impact of year-to-year discontinuities.

Sketch 2: Extended line chart

My next idea was to bring the full series onto a single plot using the same vertical and horizontal scale, then use this cleaner layout to add a severity axis that trends downward (less severe) over time as the strain becomes less deadly. I highlighted this on a secondary vertical axis using a placeholder “lethality” index, something like percentage lethality (i.e., 100% would imply everyone who catches COVID dies). I would use percentage because, to me, it gives a clearer sense of risk than something like “deaths per million infections,” for which I lack intuitive context. To highlight the main takeaways, I added an annotation and arrow for “Decreasing severity” and emphasized the winter season in color with a “Winter peaks” label. This clarifies the narrative, but at the cost of additional visual clutter.

In reference to the main purpose of our article and chart:

0. Visually interesting: Unsuccessful. This is an even more conventional chart than before and, with its extended horizontal scale, is unlikely to function well as a 1:1 feature graphic.
1. More contagious: Very successful. The spike in 2022 is unmistakable.
2. Worse in winter: Very successful. The use of color and annotation makes winter peaks easy to identify.
3. Potential for explosive growth: Very successful. The arrow in 2022 reinforces this idea, and the continuous winter 2021/2022 data makes the buildup clearer.
4. Less serious strain: Very successful. The falling lethality index, emphasized with annotation, shows that this strain is not as dangerous as earlier ones.

This is interesting: it is by far the most successful chart in conveying the four primary takeaways, yet it fails at being visually distinctive. In the click-dominated attention economy of online publishing, this is unlikely to succeed as a feature graphic. It does not matter how effectively a chart communicates data if no one chooses to view it. For my final sketch, I therefore chose to focus on visual distinctiveness.

Sketch 3: Calender chart

For this final sketch I drew inspiration from the “human experience of time” suggested in the original visualization. However, rather than treating time as a circle (or spiral) like a clockface, I reframed it as a calendar. This is a format audiences are intimately familiar with and it helps communicate just how long COVID has persisted. More than tick marks on an axis, it reinforces the day-to-day lived experience of the pandemic. Individual data points are highlighted, and by varying bubble size we see the relative magnitude or lethality of different periods. I intentionally “overplot” some days to convey the extreme growth in January 2022. I split the chart into two sections, “Covid Sick Days” (likely to be renamed “Infections” for clarity) and “Deaths from Covid.” I included graphics for both, partly for levity and partly for emphasis. I use green for sick days and dark blue/black for deaths, drawing on audience associations of green with illness and black with mortality. While circle size is not the most precise encoding, in this case it can be moderated with an appropriate scale and consideration of Stevens’ power law.

Digitally, I would experiment with showing “Infections” and “Deaths” either combined or separately, recognizing that scale adjustments would be required. Together they produce a visually rich graphic; separately they improve clarity. Replacing circle size with day-to-day shading (e.g., darker color for higher counts) would simplify interpretation but might reduce visual impact.

In reference to the main purpose of our article and chart:

0. Visually interesting: Successful. This is a distinctive and engaging plot. Featuring just the “Infections” component as a thumbnail preserves the 1:1 ratio.
1. More contagious: Successful. The spike in 2022 is clearly visible.
2. Worse in winter: Successful. Months are aligned year-on-year, making seasonal patterns apparent.
3. Potential for explosive growth: Successful. The buildup and high case load in recent data are clearly conveyed.
4. Less serious strain: Successful. The falling number of deaths suggests that this strain is less dangerous than previous ones.

This is the most visually interesting graph I have sketched and the most likely to generate clicks. While it is not quite as effective as Sketch 2 at clearly communicating all aspects of the article’s claims, with careful refinement during digital implementation it could balance visual impact and narrative clarity. As such, I will present this approach as the basis for my final graphic.

4.0 Final design

For the final design I iterated on the third (calendar) sketch to preserve what the original spiral does best: making time feel cyclical and lived, while keeping the reading task more obvious and comparisons more stable. A calendar grid is a familiar scaffold, so the viewer does not need to learn a new geometry before they can start extracting meaning. I built the chart in Python (using the july library plus custom code) and focused the narrative on two linked questions: when did the U.S. experience surges and how did severity change over time relative to case intensity. I added explicit annotations for “winter surge” and “high initial severity” to reduce guesswork and to make the intended takeaways crisp without requiring the viewer to hunt.

In terms of encodings, the calendar layout provides position for time (day-by-day), case magnitude is encoded by a transformed intensity scale (a small set of discrete levels rather than a continuous ramp to improve legibility), and severity is encoded as an additional red intensity channel. During iteration I originally plotted deaths as a second calendar, but it mostly behaved like a lagged version of cases, which did not directly communicate the article’s “less severe” claim. I therefore replaced deaths with a severity proxy and used a more selective scale so that differences in severity are visible even when case counts are high. The tradeoff is cognitive load: the viewer must keep two encoding channels in mind at once, and this inevitably reduces immediate readability compared to a single-variable chart. I accepted that trade because the assignment prompt and article context both emphasize the tension between a compelling feature graphic and a rigorous data story.

4.2 Reflection on critique-by-redesign

The process of redesigning this visualization was harder than expected. In particular our key feature that cases are worse in winter is surprisingly difficult to encode cleanly. Winter straddles the year boundary, which creates a structural discontinuity in both line charts and calendar layout, which reduces its immediate legibility. While we could have chosen to begin the calendar year in June or another non-standard month to center winter visually, doing so would likely confuse audiences who are deeply familiar with the January–December structure. This revealed that some difficulties are not purely visual but structural, embedded in how we socially organize time.

Including both infectivity and severity on the same plot also introduced a genuine design tension. Encoding only case counts simplifies the chart and makes it easier to read at a glance, but omits a central article claim (4) about reduced severity. Adding a severity metric restores narrative completeness but requires the audience to track two quantities simultaneously, increasing cognitive load. In the end, I chose to privilege narrative completeness over immediate simplicity, accepting a modest reduction in readability in order to better reflect the article’s argument.

The redesign process also made me appreciate the original graphic in its publication context. The spiral works effectively as a striking feature image paired with more conventional, easier-to-read charts within the article. Much of my initial critique focused on the spiral in isolation, but attempting to design something that is both visually distinctive and analytically clear made the tension between those goals tangible. My own effort to balance a memorable thumbnail with a rigorous data representation revealed why the original designers likely made the tradeoffs they did.