Nowadays, the search for protein-rich foods is on the prescription of many fitness and muscle-building diets. But what else are we consuming along with these proteins? The idea of these charts is to support us in making a more holistic decision about the food we eat by measuring nutritious qualities beyond the protein, along with their cost to the environment in their production.

About 50 types of food were studied, categorized into three main groups based on their source of protein: animal, dairy, and plant-based. The selection considered available data regarding their environmental cost —scarcer than nutritional values—, focus on ingredients rather than meals like pizza or smoothies, and whether they were in their raw state rather than as prepared meals. A point to comment on is that the data came from two separate sources, the US Department of Agriculture for the nutrition values and the Our World in Data for the environmental data. As a result, there was no common ID or variable, and the item descriptions were different in most cases. In these situations, they were related manually. An example of this is “chickpeas” and “chickpeas (garbanzo beans, bengal gram), dry.” To do such a study, three groups of charts were produced in Tableau: radar charts, scatter plots, and bars. For each group, different variables of the foods were intersected so the user could use several approaches and parameters to make a better comparison. The aim was to get as close as possible to tangible measures from an average person.

Radar Charts

Starting with the chart with the broadest graph, the radar chart was picked mainly because of the possibility to include as many variables as it wants. In addition, it was a way to circle the challenge of working with such different scales of measures. Counting with an axis for each value, it was possible to manage the scales for a proper comparison. However, the values with the same unit of measure, such as grams, were not modified. All the eight axes were equalized for a 100g portion of each food.

On the right side of the circle were pulled the nutritious qualities of fiber, calories, fat (lipid), and cholesterol (at the bottom). And on the left side, the environmental costs of the carbon footprint, use of land for the production, and water wasted in the process. And at the top is the protein, which was chosen to conduct the study. Initially, all the foods overlapped in the same radar, which made it hard to find patterns and isolate average measures. The decision, so, was to unfold into four radars, one for each category of food –and then was possible to express sub-categories by gradient of colors– and a fourth one with all the foods, colored only by their category color.

The following steps for these charts will be to find a way to point the label for each radar. At the moment, they are just the resulting shape. The radars were produced with an extension called LaData Viz offered by the Tableau desktop software, and despite being simple to work with, it has some limitations (especially in the free version).

Scatter Plots

Another way to demonstrate relations among various attributes of the food was to print scatter plots. The chart allows us to use colors and sizes in addition to the two axes, x and y. The solution here was to keep the protein constantly on the x-axis and vary the y and size of the points.

I was lucky to present a previous version of the scatter plots in class, and one of the feedback received was the uncertainty about the visual meaning of size. Based on that, the first chart adopted the number of servings within a 100g portion as the size since one could relate the size of portions with the size of the circle.

I believe that the advantage of this representation is that it shows that the criteria for making a holistic decision are not always so simple, and we have to consider many aspects. In that case, beef, for instance, causes a clearly high cost to the environment. However, a typical portion of it —that said, a portion that people commonly consume beef—, is enough to provide a good amount of protein. Another solution to rely on a more meaningful criterion for size could be to find a way to actually show the number of servings within the 100g in the icon. However, this could get too noisy for a two-axes chart. In the second chart, the size was used to demonstrate the food’s carbon footprint, while the axes were dedicated to the nutritious values of protein and fat, both in grams. This was an attempt to organize the type of data using the exact representation. All the cases use colors for the three categories. Another comment that came up in the class feedback was to bring images of the food instead of working with sizes. The third version is a base for that idea that I didn’t figure out how to do it yet.

Bars

The bars were used as a unit progress bar chart to mark the number of typical servings within 100g of weight and 100g of protein. The goal was to bring the measures closer to how we interact with those foods. To get 100g of protein —that would be an average amount recommended to consume daily— only with chia, for instance, one would have to ingest about 60 portions of it. It’s a straightforward chart that could complement the other ones.

Next Steps

The immediate next step is to work on better legends and distinguishable identification of the foods. Also, after reflecting on the relationship between these two different data types, I believe there is much more to explore