When you're putting together a chart, you're trying to show one of four things with the data you have: a relationship between data points, a comparison of data points, a composition of data, or a distribution of data.
1.Relationship of data.
It tries to show a connection or correlation between two or more variables through the data presented, like predicting the sales of next year according to sales data of nearly five years.
2.Comparison of data.
It tries to set one set of variables apart from another, and display how those two variables interact, like the sales to two years in each month.
3.Composition of data.
It tries to collect different types of information that make up a whole and display them together, like the search browsers that those visitors used to land on your site.
4.Distribution of data.
It tries to lay out a collection of related or unrelated information simple to see how it correlates, if at all, and to understand if there's any interaction between the variables, like the number of bugs reported during each month of a beta.
Once you understand what message you're trying to send with the data you have, it's time to select the best method for displaying that information. Different chart types cater best to different methods. For example, scatter plots are best used to show distributions, while line charts (essentially, scatter plots with a defined trend) are better suited for relationships. Pie charts do well when you're trying to communicate a composition, but make for poor comparisons or distributions .
This flowchart can help you select the best type of chart for the message you want to send.
Once you've selected the right type of chart for your data, make sure you don't do your data a disservice by forgetting some basic design tips. Kill the grid lines unless they're absolutely necessary, or at least make them subtle so they don't distract from the information you're trying to present. Make sure your chart is centered on the data you want to present, your axes are clearly labeled, and your axes have units on them where necessary, so no one has to guess or infer what you're trying to say. Remember, your goal is that anyone can pick up your chart, whether you're there to talk to it or not, and understand what information the data is trying to communicate.
(Sources from lifehacker)