Launch, grow, and unlock your career in data

February 25, 2013

Nielsen Makeover: Global recession sentiment chart alternatives


Nielsen recently published a report on economic recession sentiment, which included this summary chart:

At first glance, you might think “This is simple enough, how could Nielsen screw this one up?”  Fortunately for me, they continue to make a mess of even the simplest charts. 

This chart makes North America and Europe negative sentiment look so much higher than any other region or the global average.  However, that’s only because the scale does not start at zero, nor do they call that fact out.  Look at Latin America.  Since when is 53% about 1/4 of 75%?

I wanted to see how far off their chart was so I built a simple bar chart in Tableau as a first alternative.


This is a very different perspective.  Europe, Middle East/Africa and North America are all pretty close.  And the change by quarter does not look nearly as severe.

A simple bar chart is an effective way to present this data, and there are other alternatives that work as well.  Let’s consider two slope graphs. 

The technique used to build these slope charts is very similar to the one I outlined in my previous blog post.  Notice that the lines are colored by whether the sentiment increased (got worse) or decreased (got better).  For this coloring, I used a technique recommended by reader Santiago Restrepo in which we calculate the quarterly change as a discrete measure.

First, you need to create a calculated field:


By default, the table calculate is set to compute along the Table (Across).  For the slope graphs, we want the calculation to calculate by Quarter.  To do so, click on the Default Table Calculation link on the Calculated Field window and choose Quarter.


The first slope graph looks at all Regions together:


The second option splits out each region across the table.


Which of the three options do you prefer?  The bar chart or maybe one of the slope graphs?

Do you want to build something to share?  Download the data here and/or the Tableau workbook here and leave a comment with a link to your work.


  1. Love the first slope graph. You can easily compare countries, quarters, and quarters by country.

    Personally, I feel like it is more difficult to estimate the values on the second slope graph. Maybe adding a '% change' label could be a nice addition.

    The bar graph is nice.

  2. I think your slope graphs would benefit from not starting the axis at zero, less wasted white space and more precision. Your bar chart and second slope graph would also be better with a meaningful sort. I can understand putting the Global Average first or last, but alphabetical is rarely a useful sort.

    Another option to consider is an arrow chart, maybe something like:

    Also, the screen shot of your formula for "Quarterly Change", may get you the results you are looking, but is less than ideal. A better formula would be something like:

    IF LOOKUP(SUM([Percent]),LAST()) > LOOKUP(SUM([Percent]),FIRST()) THEN "Increase"
    ELSEIF LOOKUP(SUM([Percent]),LAST()) < LOOKUP(SUM([Percent]),FIRST()) THEN "Decrease"
    ELSE "Same" END

    1. Thanks for the comments Joe.

      I intentionally kept the axis at zero on the slope graph so that the slope was accurate. It was definitely a consideration though. I go back and forth on this with slope graphs.

      I think the slop graph works better than the arrow chart in this case. I feel like direction (falling or rising) makes it more intuitive. It'd be even better if I could put arrows on the ends of the lines in the slope graph.

      I sorted the bar chart the same way was the Nielsen version so that I could compare the two. A more meaningful sort would be good, probably by Q4 2012 sentiment increasing.

      Thanks for the tip on the formula. I'm by no means an expert on calculations, so I appreciate the feedback. No one knows better than you which functions are more efficient than others.

    2. Thanks for the reply Andy,

      I am interested in why you think having axis of the slope graph include zero makes the slope accurate. I find that by not including zero in a line chart enables greater accuracy through greater detail.

      When dealing with a bar chart, the axis should include zero, because the viewer is comparing the bar length, therefore not starting at zero is deceptive to the way we perceive bar graphics, because the length is no longer comparable.

      When dealing with a line chart, the distance from zero is no longer the primary comparison. The primary comparison becomes the slope between points, and not starting at zero can be very beneficial, especially so when dealing with a slope graph. It is beneficial because it zooms the graphic, enabling greater detail in the slope to be seen.

      Here is a side by side comparison:
      1. Which enables greater detail to be seen?
      2. Why do you think not including zero makes the chart inaccurate?

      A line type chart enables relative comparisons, and starting at zero shows less detail.

      Here is how I would make a slope graph of this data:

      Why do I think the the Arrow chart is better than the slope graph for this data set?
      1. Can be sorted
      2. Uses less charting space
      3. Displays the difference as well as the start and end
      4. Less worry about label overlap (no need for manual label position)
      5. Confidence with larger click target when hovering for a tool-tip or selecting a mark for interaction.
      6. Fits well with this data source because there is a limited number of items

      An arrow chart can become less useful than a slope graph when there is a greater number of items, because we would loose the single overview, but that is not the situation of this data set.

      From my perspective an arrow chart is like a Gantt chart (floating bars) with a mark and color to denote direction. What do you find non-intuitive about an arrow chart?

      In regards to intuition, what do you think of preattentive attributes? like from Stephen Few, or rankings of data encoding methods such as Cleveland's graphical features hierarchy, as seen in

      My biggest negative about the arrow chart is that it is not immediately clear what part of the arrow mark is the end point, is it the tip, the base, or the middle of the triangle mark. In this case, it is the middle of the mark.

      Why do you think putting arrow heads on the slope graph would make it better? I think adding additional marks to the line ends would further confound the overprinting that is inherent with a slope graph, you already have color and slope to denote direction.

    3. In my opinion, if you exclude 0 on the slope graph you're committing the same sin as the bar chart. You're distorting the differences.

    4. Andy, Can you please help me see how you come to the opinion that by not including zero the differences are distorted?

      for example:
      How is the "Not Include Zero" chart distorting and in what way?

      I would like to better understand you opinion.

  3. Thanks for another great example where a slope chart tells the story in a much more intelligible way than simple bars