Ugly graph

Shown here is a graph that displays statistics on car sales trends. While the information is there, its not exactly readable and I would say an eye-sore, you have to peer in and start matching colours and signs to understand whats going on in the image.

What would be the best way to display this data beautifully and make it easy to understand?.

  • The graph shows the sales of each car manufacturer in thousands per month. I would like to display that on a graph I've thought about using colour coded circles but that was again overlapping with other data that made it look like a Venn diagram. Bar charts wont work either and so wouldn't pie charts. – Jezza Jan 21 '14 at 14:01
  • 1
    Do you need to compare manufacturers to each other or just their own previous performance? – bendataclear Jan 21 '14 at 14:19
  • @bendataclear: I would say retrieve sales data of a country, notice sales data trends. Compare year-on-year sales for each manufacturer and present this as a percentage change. – Jezza Jan 21 '14 at 15:11

Your graph is not scaled properly due to the presence of outlying data (Maruti). I would advise omitting or putting that data into a separate graph so that you can fit the rest of the data to fill up the graph.

At least 80% of the area should be occupied by data. Also, either omitting the data points or making the data lines thinner would make the graph cleaner.

  • I thought of omitting it but that would only address the smaller elephant in the room. I would like to know if those colour codes of each manufacturer be omitted without removing its dependency. – Jezza Jan 21 '14 at 14:05
  • 80% of the area should be occupied by data. That point made me choose this as the answer. – Jezza Jan 22 '14 at 10:28

Display fewer data series would be my first impulse (say top five only), but that doesn't address the primary question that should be addressed: What data insights are you trying to extract?

  • Sales patterns throughout the year? In that case you can probably merge several/all Manufacturers
  • Proportion of market share between manufacturers? Then perhaps you should change the chart to a 100% stacked type.
  • Connecting events (new cars models, price deals, etc) to surges or drops in sales? Then you should look to annotate the chart with these events.

Until you can be clear about the insights you want to make, it's difficult to choose the appropriate chart tool to use.

  • Data insights to extract: display every manufacturers monthly sales data in a graph. How do I go about it? – Jezza Jan 21 '14 at 14:08
  • Sorry, I'm asking what is the purpose the chart will be used for? What will the persons reading it intend to understand is happening or is not happening? – Fractional Jan 21 '14 at 14:13
  • The chart will be used to show which of the manufacturers have had a good year based on sales. People reading the graph will take a look at this chart and analyse which manufacturers enjoyed a good year and which ones didn't?. What is the scenario of car sales throughout the year. I hope I was clear this time. – Jezza Jan 21 '14 at 14:16
  • Well a 'good year' is probably subjective and different for each manufacturer. If a small manufacturer doubles their sales they'll probably remain near the bottom of your chart. – Fractional Jan 21 '14 at 14:19
  • So maybe compare year-on-year sales for each manufacturer and present this as a percentage change. So April 2014 compared April 2013, etc. – Fractional Jan 21 '14 at 14:20

Here is a good example for removing clutter in a chart:


Also, i would suggest to use a logarithmic scale for the y-axis so the lines are more spread out.

  • I wanted to display that link in my question. While the points presented are good I still am not able to think how do I implement these in my graph? – Jezza Jan 21 '14 at 15:13
  • I would start with removing the borders, lighten the gridlines, lose the datapoints and reduce the line width of the series. It will be less cluttered that way. – defreek Jan 21 '14 at 16:17

Some points are:

  • what you display is the raw data, not the trend. To make trend you should perform some statisitical processing over the data. Trend lines are really perceived easily, but you'll lose some leftmost and rightmost data points
  • divide all the manufacturers on three or four ranges. The ranges can be not equal. This will simplify overal data view, still the general idea will be conveyed
  • break the Y-axis to fit the data

enter image description here

  • This just gave me a new idea. I'm trying it out now, will let you know if it solves my problem. – Jezza Jan 21 '14 at 15:58

With that amount of data you'd likely want to consider implementing a way to "switch" on/off options and maybe display only five or six (top trending) by default. Put in a subtle icon (flame or something) next to the top 5 indicating that they're the top trending to distinguish them from the others below it on the key.

enter image description here

Use hover overs to display more meta data as well.

  • Would this be as a jquery graph or some other tech used online? That option can't be used while sharing data with many people especially in my government, most people want it to be shown as a graph and wouldn't like the idea of interactive graphs. Maybe your solution can be used if I display it using a gif file showing the hover overs. Still the people viewing this data aren't tech savvy. Plus showing it as a gif would pose a new problem: for viewing a particular data I must watch the entire gif and record data instantly. – Jezza Jan 21 '14 at 14:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.