14 Charting Tips: Tip #5 Number of Data Points

This is the 5th post in the series, 14 Tips to Present Awesome Charts.

After choosing the chart type, giving the chart a suitable title and presenting the key finding you now have to decide on how many data points should your chart have.
What is a Data Point?
If your pie chart has 10 parts then it has 10 data points. If the bar chart has 4 vertical bars, there is 1 bar for each of the 4 data points. If you are making a one year chart of the stock price of Infosys, then you will need 365 data points (actually you would need somewhere around 310 because there are 52 Sundays plus some holidays when the share market is closed).

You must we wondering why worry about such a small thing as a data point. Why devote one complete post on it. The reason is, number of data points deserves that kind of attention.

How do you decide on the correct number of data points?
As you know, a chart is used to prove or disprove a point to the audience. In order to do the analysis, you might need 20 data points but to prove the point in the presentation after the analysis you might only need 10. So when you 'present' your chart you must have only 10 data points. But you generally err on this front and display every piece of data you have. You go with more data points than needed. What happens when your chart has too many data points?

1. The more the data points, the more complex the chart becomes.

2. Audience understanding is inversely proportional to the no. of data points your chart has. This means the more the data points, the lesser the understanding in a given time. If the audience has half a minute to understand your chart, the lesser the no. of data points, the better it is.
Example 1: KK Consultants (name changed) are making a presentation to the new employees of their organization. They are sharing how their organisation has grown leaps and bounds in the last three decades. This is how the chart looks (I have jazzed it up a bit, the original looks really cluttered):
Perfectly acceptable. But think of the new employee looking through this and trying to read 15 data points. The key message here is 'that our organization has grown leaps and bounds over the last three decades. To make this argument, you might need to analyze 15 data points, but to present you can manage with only 4. Take a look at this new chart: Much better. For an audience to understand 14 data points takes time and does not add much value. By reading 14 data points, it is tough to draw much conclusions. Humans cannot mentally analyze so many data points. But with 4 data points, the audience gets the key message far more easily and far more powerfully.

Example 2: CEO of Tyger Toothpaste wanted to know what paste people use in the city of Bangalore. He appointed Mr. Jay to conduct a survey. Jay asked 1000 people which brand of toothpaste they use. Here is the result of the 15 brands people use:
How will Jay make the chart? Remember what we discussed in Tip 2 on 'How to choose a chart?' Jay is breaking down the number 100 into smaller parts and hence he can use a pie, a Bar or a Stacked Bar Chart. But a pie almost always looks better. This is how Jay's chart looks like:
Nothing wrong. This is how everyone generally does it. Can you do better? Think for a while before checking my answer. The objective of this chart is to inform the CEO about the brand preference for toothpastes in the city. He would like to know where his brand stands versus competition. Try making this chart like this: Jay needs to tell the CEO that Tyger stands 4th in the market and total share of these brands is 72% (this may well be the key message). By showing all the 15 brands Jay is adding to the clutter. It is better to focus on a few main brands in the presentation and give a print out of the complete data while you are presenting this chart.
3. Choose equidistant data points when presenting data over time. I would explain this point with the help of an example.

Example 3: This chart is from the American Heart Associations presentation. It talks about the Heart disease mortality trends. Look closely at the chosen data points; 1979,1980, 1985, 1990, 1995, 2000 and 2005. While 1980 to 2005 are at 5 year gaps, there is also an unwanted 1979. Visually all data points look equally far away from each other in time. But they are not. This situation should be avoided. Always choose equidistant data.

To summarize, remember these two tips on data points:

1. Keep the number of data points in a chart to as less as possible2. In case of presenting data over time, choose data at equal time gaps

--
Disclaimer: All charts have been used for educational purposes. It is not meant to comment on the working of any organization. They can be removed in case of any objection.