Five steps to let the young ones shine
Nice example of visual story-telling in the FT

Not following direction or order, the dieticians complain

At first glance, this graphic's message seems clear: what proportion of Americans are exceeding or lagging guidelines for consumption of different food groups. Blue for exceeding; orange for lagging. The stacked bars are lined up at the central divider - the point of meeting recommended volumes - to make it easy to compare relative proportions.


The original chart is here, on the website.

The little icons illustrating the food groups are cute and unintrusive.

It's when you read further that things start to get complicated. The last three rows display a flipping of the color scheme, with orange on the right, blue on the left. Up to this point, you may understand blue to mean over the recommended value, and orange is under. Suddenly, the orange is shown on the right side.

The designer was wrestling with a structural issue in the data. The last three food groups - sugars, fats and sodium - are things to eat less. So, having long bars on the right side is not good. The orange/blue colors should be interpreted as bad/good and not as under/over.

The problem with this design is that it draws attention to this color flip - that is to say, it draws attention to which food groups are favored and which ones are to be avoided. This insight is actually in the metadata, not what this dataset is about.

In the following chart, I enforce the bad/good color scheme while ignoring the direction of good. The text is adjusted to use words that do not suggest direction.


Dieticians are probably distressed by this chart, given that most Americans are lagging on almost all of the recommendations.

In a final edit, I re-ordered the categories.




Feed You can follow this conversation by subscribing to the comment feed for this post.

Colin Fredericks

Ugh. You can't fix this graph because the data is bad from the start.

Eating more of something or less of something should not be described as absolutely good or bad unless that thing is not a food. Eating closer to the recommended amount should be good, and eating far away from the recommended amount should be bad, otherwise a recommendation to eat a specific amount makes no sense.

Also, "percent of population above/below" tells us nothing about how close those people are to the recommended amount. It's clear in the original graph that a large percentage of people get too much salt, but how much more?

The worst thing about this chart, though, is that the percentages add up to 100% in each row, which means that no one is eating the right amount of anything. This is either incredibly dire or total BS, and my money is on the latter.


CF: Can't argue with that. You're arguing against all "proficiency" metrics. It's a pretty complicated issue - having one standard for all Americans also seems wrong, but it's tough to communicate a highly-nuanced recommendation for different slices of the population.


I don't see a value in making the bars float around a centre, since they're all 100% in total length. I think the chart should be stacked bars adding to 100%.

The comments to this entry are closed.