Case Study 2: The New York Times 2020 COVID Small-Multiple Map

Case Study 2: The New York Times 2020 COVID Small-Multiple Map

During the pandemic, The New York Times published a single recurring image that may have done more work than any other chart in recent memory: a 50-state small multiple of COVID case trajectories, updated daily, in a layout that let readers see their state's trajectory in context without losing sight of the national picture. It is a master class in small-multiple design at scale.

The Situation

By April 2020, the COVID-19 pandemic had reached every U.S. state, and the public was desperate for information about how their state was doing relative to others. Was the outbreak under control in Oregon? Was Texas accelerating? How did Florida compare to California? Were any states flattening the curve? The questions were urgent, specific, and relentlessly local — every reader wanted to know about their own state, but also about the handful of other states they had personal or professional reasons to care about.

The standard tools for answering these questions failed at this scale. A single national chart of total U.S. cases hid the state-level variation. A scatter plot of cases by state showed snapshots but not trajectories. A line chart with 50 overlapping state series was unreadable — the "spaghetti chart" problem. A table of current case counts by state gave numbers but not shapes. Each of these representations answered some part of the question but failed at the core task: letting a reader see their state's trajectory in the context of all 50 states simultaneously.

The New York Times graphics team, which had been producing a daily-updated pandemic dashboard since early March, needed a visual that could handle this multi-dimensional question in a form that millions of readers — most of whom were not data-visualization experts — could read in seconds. Their answer was a small-multiple display: 50 tiny line charts arranged in a grid, each showing the case trajectory for one state, with the reader's attention drawn to selected states via color and direct labeling.

The small-multiple COVID display was published in dozens of variants over the course of 2020 and 2021 — sometimes as cases per capita, sometimes as deaths, sometimes with a highlighted subset of states, sometimes in a geographic arrangement that matched the U.S. map, sometimes in a ranked arrangement by current case rate. The core design — 50 small panels, shared time axis, consistent visual grammar — remained constant across the variants. The design became recognizable as "the NYT COVID map," and it was adapted by many other outlets including the Washington Post, Reuters, and regional newspapers.

This case study is worth examining because the NYT COVID small multiple shows the technique working at scale, on a high-stakes topic, for an audience of tens of millions, updated daily for over a year. Every design decision the team made was tested by millions of reader interactions over an extended period. The decisions that survived are the decisions worth learning.

The Data

The underlying data was a daily-updated record of COVID-19 cases and deaths for each U.S. state. The NYT team sourced and verified the data themselves — in fact, the NYT maintained one of the most widely-cited public COVID case databases, which became a primary source for other researchers and news outlets. For each state and each day, the data included:

Cumulative cases (total confirmed cases since the start of the outbreak)
Daily new cases (cases reported that day)
Seven-day rolling average (new cases smoothed over a week to remove reporting artifacts)
Per-capita measures (cases per 100,000 residents, for fair comparison between states with different populations)
Cumulative and daily deaths in parallel structures
Testing rates and positive test rates (later additions)

The most-used measure in the small-multiple display was the seven-day rolling average of new cases per capita. This measure removed the reporting noise (weekend dips, reporting delays) that would have made raw daily counts chaotic, and the per-capita normalization let a reader compare California's trajectory to Wyoming's without being misled by their enormous difference in population.

The data was updated daily — in many cases, hourly during the most active phases of the pandemic. This meant the small-multiple display was a live visualization: every time a reader visited the NYT pandemic page, the chart reflected the latest data. The design had to work not just for one snapshot in time but for continuous updates over a multi-year period.

The Visualization

The NYT COVID small multiple went through several versions, but the core design elements were consistent.

Layout: 50 panels in a grid arrangement. In the geographic variant, the panels were arranged in a rough approximation of the U.S. map (Washington state in the upper-left, Florida in the lower-right, other states in between). In the ranked variant, the panels were ordered by current case rate (highest at the top). Both variants were used at different times, and both were effective for different purposes.

Panel design: Each panel was a small line chart with: - X-axis: Time, running from March 2020 to the current date, shared across all panels. - Y-axis: Seven-day rolling average of cases per 100,000 residents. The y-axis was typically shared (same scale across all panels) so that magnitudes could be compared directly. - The line: A thin dark line showing the state's trajectory. - Fill: Often a subtle fill under the line in a muted color, to make the shape of the trajectory more visible. - Panel title: The state's name or two-letter abbreviation, placed above the panel in a small font.

Shared elements across the figure: - Figure title: An action title describing the current state of the pandemic. "Coronavirus Cases Are Rising in Most States." "New Cases Continue to Fall Nationally." The title changed as the overall trend changed. - Figure subtitle: The date, the metric (e.g., "Seven-day average of new cases per 100,000 residents"), and the scale information. - Source attribution: "Source: State and local health agencies (via The New York Times)." - Color palette: Mostly grayscale for the line charts, with one or two accent colors used sparingly to highlight specific states when relevant.

Highlighting: In many variants, the display used the "highlight strategy" from Chapter 7. A handful of states that the article was specifically discussing were drawn in a bright color (perhaps red for "high case rate" or blue for "trending down"), while the other states were drawn in muted gray. The highlight drew the reader's eye to the states the story was about while preserving the full context of all 50 states.

Direct labeling: Where states were highlighted, their names were typically labeled directly in a larger font or with a visual callout, rather than requiring the reader to hunt through the grid for that state's panel.

Reference lines: Some variants included horizontal reference lines at meaningful thresholds (e.g., 10 cases per 100,000, which was often cited as a threshold for "controlled spread"). These reference lines appeared in every panel and let the reader see whether a state's line was above or below the threshold at a glance.

Time axes: The time axis was shared across all panels, so that the reader could compare any two states at the same point in time by looking at the corresponding x-coordinates in both panels. Events like "first case," "first death," "first vaccine" were implied by vertical reference lines in some variants.

The Impact

The NYT COVID small multiple was one of the most-viewed charts of 2020–2021. The NYT pandemic page received tens of millions of unique visitors at peak periods, and the small multiple was a centerpiece of that page. A reader visiting the page saw the grid of 50 state trajectories and could identify their own state's trajectory in a few seconds — most state panels had recognizable abbreviation labels, and the geographic variant arranged them in a layout that matched mental maps.

The display worked, in the sense that readers could extract meaningful information from it in seconds. Specific reader behaviors the design enabled:

1. "How is my state doing?" A reader could locate their state's panel (by geographic position in the map variant, or by reading the labels in the ranked variant) and immediately see whether the state's line was rising, falling, or flattening. The shape of the trajectory was the primary information, and the shape was visible without needing to read any numbers.

2. "How does my state compare to my neighbors?" In the geographic variant, adjacent states were adjacent in the display, so a reader in California could see Nevada, Oregon, Arizona, and the Pacific Coast states at a glance. The spatial arrangement made geographic comparisons trivial.

3. "How does my state compare to the worst-performing states?" In the ranked variant, the worst states were at the top of the grid. A reader could see where their state was in the ranking and how far their state was from the best or worst performers.

4. "How does my state's current trajectory look relative to the past?" The panel showed the full trajectory from March 2020 to the present, so a reader could see whether the current rate was higher or lower than previous peaks, how long a surge had been going, and whether any earlier declines had been sustained.

5. "What is the national picture?" A reader scanning the full grid could see whether most panels were trending up or down, whether the pattern was regional or diffuse, and whether the "national" story was consistent with their state's experience. The small multiple did the work of a national summary by letting the reader aggregate visually across panels.

Five different questions, one chart. This is what good small multiples do: they let the reader answer multiple related questions from a single well-designed layout.

The chart's influence extended beyond the NYT. The Washington Post, Reuters Graphics, The Guardian, regional newspapers, and state health departments all produced variants of the 50-panel small multiple during the pandemic. The pattern became a standard way to visualize state-level pandemic data, and the general public became familiar with reading small-multiple displays as a result. A generation of news readers was trained on small multiples by the pandemic coverage, which in turn has raised expectations for other kinds of visualizations.

Why It Worked: A Theoretical Analysis

The NYT COVID small multiple succeeded because it applied principles from this chapter consistently and at scale.

1. Enforced comparison across states. Every panel used the same chart type (line chart of cases per capita over time), the same y-axis scale (usually), the same x-axis scale, and the same visual design. A reader could compare any two states by looking at their adjacent panels and immediately see which was higher, which was rising faster, which had peaked earlier. The comparison was built into the visual structure, not computed by the reader. This is the core power of small multiples from Section 8.3.

2. Shared x-axis enabled temporal alignment. Every state's trajectory was plotted on the same time range, so comparing any two states at the same date was as simple as looking at the same x-coordinate in both panels. No mental rescaling was required. This is the alignment principle from Section 8.2, applied at figure scale.

3. Consistent design created visual unity. Every panel used the same thin line, the same muted fill, the same small title, the same axis treatment. The consistency meant the full 50-panel grid felt like one figure, not 50 disconnected images. The Gestalt principle of similarity was operating at maximum strength.

4. Geographic arrangement matched mental maps. When the panels were arranged in a rough approximation of the U.S. map, the spatial arrangement carried additional meaning: the reader could use their knowledge of U.S. geography to navigate the grid. This is a small-scale version of Marey's spatial alignment principle — the layout of the figure mirrored the real-world structure the data was about.

5. Highlight strategy focused attention without losing context. When the article was about specific states, those states were colored and labeled directly while the other 49 states remained visible in muted gray. The reader's attention was drawn to the highlighted states, but the context of "what are the other states doing?" was preserved. This is the highlight strategy from Chapter 7, applied to small multiples.

6. Consistent visual grammar across variants. The NYT produced many variants of the small multiple — geographic, ranked, filtered, annotated, with different metrics, with different highlights — but all of them shared the core design language. A reader who had learned to read one variant could read any variant, because the visual grammar was stable. This is the institutional consistency principle from the NYT case study in Chapter 7 (and the Apple Health case study in Chapter 6).

7. The action title framed the current story. The chart was not a neutral display of 50 state trajectories. It was a reported story, with an action title that stated the national finding: "Cases Are Rising in Most States" or "The Pandemic Is Declining Nationally." The action title told the reader what conclusion to reach from looking at the grid, and the grid provided the evidence. This is the typography discipline from Chapter 7, applied to a multi-chart figure.

8. Daily updates kept the display live. The chart was updated every day (and sometimes more often), so a returning reader saw a continuously evolving story. The design had to work not just for one snapshot but for a series of snapshots over time. The NYT team's discipline was in maintaining the design quality across hundreds of daily updates over a multi-year period.

Complications and Limits

The NYT COVID small multiple was praised but also criticized, and the critiques are worth understanding.

The shared y-axis hid state-level detail. For states with very low case rates, the trajectory appeared as a flat line near the bottom of the panel, because the y-axis was scaled to accommodate the high-case states. A reader in a low-rate state saw their line as "essentially zero" even when the state's trajectory was meaningfully moving. The alternative — free y-axes — would have made each panel's pattern visible at its own scale but would have sacrificed the magnitude comparison across states. The NYT team chose shared axes to preserve comparability, at the cost of hiding detail in low-rate states. This is a legitimate trade-off, but not everyone agreed with the choice.

50 panels is close to the upper limit of readability. A reader scanning 50 panels has a lot of work to do. Some critics argued that the display was visually overwhelming, especially when the viewer was looking for a specific state they did not know how to find. The geographic arrangement helped, but not everyone knows U.S. state positions well enough to navigate a map-like grid. The ranked arrangement helped, but only for states the reader knew were high or low. For the large number of readers who did not immediately find their state, the grid was a search problem.

The shared design across variants could be confusing. When the NYT published different variants of the small multiple (geographic, ranked, filtered), a reader returning to the page might be confused by the changed arrangement. The visual grammar was stable, but the specific layout could change, and that change sometimes disoriented readers who had mentally bookmarked their state's position in the previous variant.

Small panels limited the ability to annotate. With each panel perhaps 1.5 inches wide, there was no room for per-panel annotations, callouts, or explanatory text. All the narrative work had to happen at the figure level (via the title and subtitle) or via accompanying prose. This is a general limitation of small multiples: individual panels cannot carry much text.

The pandemic-specific context does not generalize cleanly. The NYT small multiple worked partly because readers were intensely motivated to understand the pandemic and were willing to spend more time with the chart than they would with a typical news graphic. In less urgent contexts, the 50-panel density might be too much for the available attention budget. Small multiples at this scale are best suited to high-engagement, high-stakes contexts.

Lessons for Modern Practice

The NYT COVID small multiple offers several specific lessons for practitioners who will never work on a pandemic visualization but will work on multi-group comparison problems.

Shared axes enable comparison; free axes enable pattern visibility. The choice between the two depends on what you want the reader to see. Shared axes answer "which group has the highest value?" Free axes answer "what is the shape of each group's pattern?" Pick deliberately, and label the choice clearly so the reader knows what kind of comparison is possible.

Geographic arrangement carries extra meaning when it matches mental maps. If your groups have a natural spatial relationship (states, countries, regions), arranging the panels to match the geography lets the reader navigate using prior knowledge. This is a small improvement in isolation but it compounds across many readings. When geographic arrangement is not meaningful (business product lines, customer segments), a different arrangement (by value, alphabetically, by category) may work better.

The highlight strategy scales to very large small multiples. With 50 panels, the reader cannot pay equal attention to every panel. The highlight strategy solves this: draw most panels in muted gray and highlight the two or three that matter for the current story. The reader's attention goes to the highlighted panels, but the context of the other panels is preserved. This works whether you have 10 panels or 100.

Consistency across a body of work compounds benefits. The NYT team produced dozens of variants of the small multiple over a multi-year period. The consistency across variants meant that readers who had learned to read one variant could read any variant. If you produce charts in a series — monthly reports, weekly updates, sections of a longer document — consistency across the series pays off in reduced cognitive cost for the reader.

Update discipline matters for live visualizations. A chart that updates daily for a year is effectively 365 different charts, and maintaining quality across all of them is a discipline in itself. Design for sustainability: make the update process automated, make the defaults robust to unusual data, make the chart degrade gracefully when something is wrong. Short-term beauty that breaks after three updates is not as valuable as modest beauty that holds up across hundreds of updates.

Small multiples can carry the national story. A reader looking at 50 state panels can see the national picture by scanning the grid and aggregating visually across panels. The national story emerges from the pattern of the individual panels, without any national-level summary chart. This is a subtle and powerful property of well-designed small multiples: they let the reader see both the individual-group detail and the aggregate pattern in a single figure.

Know when 50 panels is too many. The NYT display used 50 panels because U.S. states are a natural unit and readers specifically wanted state-level information. With 200 panels, the display would have been unreadable. With 500 panels, impossible. The limit depends on the reader's engagement level and the size of the available display, but somewhere in the 20–80 range is the practical limit for most small multiples. Beyond that, consider aggregation (group the groups into higher-level categories) or interactive filtering (let the reader choose which panels to display).

Discussion Questions

On shared vs. free axes. The NYT team chose shared y-axes for their state small multiples, which hid detail in low-rate states but preserved comparability. Would you have made the same choice? Under what conditions would you switch to free axes?
On the highlight strategy at scale. The NYT used the highlight strategy to draw attention to specific states while keeping the other 49 visible in muted gray. Could you use the same technique in your own work with fewer groups — say, 10 product lines where one is the "current quarter's story"?
On geographic arrangement. The map-like arrangement of state panels carried meaning because the reader already knew U.S. geography. In your own work, do you have groups with a natural spatial or ordinal structure that could be used for panel arrangement? What would you lose by arranging them alphabetically instead?
On the sustainability of daily updates. The NYT chart was updated daily for over a year. What practices would you need to put in place to sustain chart quality over that many updates? How do you design for the long tail of weird data conditions?
On reaching an audience of non-experts. The NYT team's audience included tens of millions of non-experts. The small multiple worked for them because the individual panels were simple line charts — a familiar form. Would the same design have worked if the panels had been unfamiliar chart types? What does this imply about the chart-type choice in small multiples?
On the 50-panel limit. The chapter says somewhere in the 20-80 range is the practical limit for small multiples. What determines the upper limit in your specific context? Is it screen space, reader patience, data granularity, or something else?

The NYT COVID small multiple is not a unique innovation — the small-multiple technique existed before the pandemic, and other outlets produced similar visualizations. What makes it worth studying is the scale at which the technique was applied: 50 panels, updated daily, viewed by millions, for over a year, in a high-stakes context where getting it wrong had real consequences. The design held up. The principles worked. The same principles — shared axes, consistent design, highlight strategy, meaningful arrangement, action titles — will work for your problems at your scale, even if your audience is your team instead of the nation.