The value of the data is mainly reflected: the graphical way to render the data

Article Description: the value of the data mainly reflects: the graphical way to render the data.

In usability testing, the value of data is primarily in support of test results, in other words, if we can't communicate well with others, the value of our data is very limited.

It's a simpler way to render data in tabular form, but in order to better interpret the data, we need a graphical way to render the data.

Typically, the types of data graphs we use are as follows:

L Bar Chart (histogram) column or bar graphs

L Line Graphs

L Scatter fig. Scatter plots

L Pie chart Pie charts

L Stacked bar graph Stacked bar graphs

The most common of these should be bar (columnar) graphs, in the usability test report, we often use this data rendering form, in general, Bar (columnar) graph will be used to render user characteristics, task completion rate, task completion time, AB test comparison, and so on, we first on this type of data graph using the principle of discussion.

When writing usability test reports, we found that bar charts are especially good for discrete projects and categories, such as tasks, participants, designs, and so on, especially for data comparisons, which are very intuitive in terms of the length of a graphic presentation. But there is a premise that the axes must start with "0". If you do not, it looks as if the length of the column has been manipulated artificially. This is part of the subjective factor that we must avoid.

Then, the maximum value of the chart axis is best not to be higher than the theoretical maximum, which makes the chart unreasonable. For example, the highest rate of completion is 100%, when we draw a bar chart, the completion rate coordinates using 120% such marks not only meaningless, but also makes the data appear not rigorous. At the same time, too much emphasis on the accuracy of the data is not necessary, in practice, we found that in the case of no correlation coefficient, the whole number is the best situation, for example, the completion rate of 100% calibration for everyone will feel awkward.

Third, when the bar chart presents the participants ' data averages (time, ratings, etc.), it involves the concept of data variability and confidence intervals, and if we can graphically represent the variability of the data by increasing the errorbars, we can make our data easier to understand. , but also reflect the value of our data.

Figure: Standard column chart, error bars representing confidence intervals for 95%

Four, considering the diversity of data users, we should pay special attention to the way of data expression, such as, for too long title also want to use the level of way to show, to avoid the user side head to watch; for the color of the data bar to use contrast more intense color, to avoid color blindness of the user to create obstacles; for tables with large data It is suggested that the presentation be expressed in a classified manner to prevent the table from carrying too much information and to affect the presentation effect.

Finally, we say a question that we all care about: How to judge whether you should use a column chart or a line chart.

We often encounter such a situation, in the presentation of a group of data, think the column is good, line chart can also, then, how do we judge it? There is a simple way to assume that you need to use a line chart, and then we ask ourselves a question: does connecting to a data point make sense?

In other words, if there is no data in these locations, can we add data to increase the persuasive power of data?

If not, use a column chart.

So as long as the rational use of the above principles, we can not only make beautiful graphics to make boring data easy to understand, but also to increase the persuasive data, reflecting the value of data.