Common mistakes when plotting Graphs. How to use line graphs and bar graphs effectively.

Learn by comparison – Features of 4 typical graphs

In this story, I will introduce how to select and create the type of graph, so that everyone can effectively understand, what you want to express with data at a glance.

This article helps you decide when you think which graph should I used, or which graph is best to represent your data.

An important factor for effective graphs is that your graph should do one thing and one thing best.

We commonly use this term in programming, known as the Single responsibility principle.

Similarly, regarding graphs, we’ll use the terminology "One claim for one graph".

This is an effective principle when you are creating graphs or reviewing someone else’s graphs.

Let’s look at some problems that most people do when representing data.

What went wrong with this graph?

The following graph is the graph that is widely used during the proposal for market competition. The presenter just wants to explain one claim in this graph that is –

Claim: "Internet usage for home consumers is higher on weekends than on weekdays".

Looking at this graph, we can pull out various problems. Because the answer isn’t to the point, and the graph cannot withstand its original claim. There are 3 problems in this graph. Can you spot the problems?

Figure 1: Internet usage rate by day of the week. Image by Author

Now, let’s learn about the following problems with this graph and why it fails to make a strong claim.

Problems

1 Isn’t the graph just for 2020? Is it the same trend every year?

2 What is "usual 8 weeks"? It seems like it takes the data from an arbitrary selection of sets.

3 Since the scale on the vertical axis starts from 40%, there is a plunge between Friday to Sunday. But it can be also be said that there is not much difference between 52% and 57% in the first place if the scales are set properly.

"One claim for one graph" depends on how the graph is being used.

Before talking about the correct answer, let’s talk about how this wrong graph was made.

The data to represent the "usual 8 weeks" is taken from mid-May to early July for regular weekdays. This range is less affected by New Year’s holidays and summer vacation. That is the reason fixed-point observations of those eight weeks each year were taken. As there was a spike in internet usage on Saturdays and Sundays, it makes a hypothesis that we can consider this as the annual trend.

But in the principle of "one graph makes one claim easy to understand", which is the royal road of presentation. The 40% is chosen on the axis scale of the graph, so that difference between weekdays, Saturdays, and Sundays stand out. And that is the first trick that a presenter tries to catch the eyeballs of the audience by exaggerating from bold claims.

Now, please see the graph below.

Figure 2: Internet usage rate by day of the week. Image by Author

The presenter also knows that it’s possible to exaggerate by changing the scale of the graph, so the first thing that can be pointed out is – it would not make a big difference in the home internet usage – when the minimum scale is 0% on the Y-axis. Certainly, if you create a scale in this way, you will not be so concerned about the high utilization rate on Saturdays and Sundays.

To make this graph easier to understand, I haven’t started the scale from 0%, but it seems that the activity is generally low on weekdays and the utilization rate is higher on Saturdays and Sundays than on weekdays. Also to make it make a claim better, it is better to show YOY results instead of just one year for a better understanding of the general trend. Following is a much better representation of the graph that makes effective claims.

Also note that, when we add comparison, we don’t need the header captions and the graph can involve when new data is added.

Figure 3: Internet usage rate by day of the week. Image by Author

Excessive data processing is dangerous. Let’s learn the basics first

Sometimes, it is possible to select only the data that is convenient for you and use it for the presentation. But even if you win the hearts with a wonderful presentation at that time, the data used is almost a lie.

However, once it is known later that the data was in fact a lie, it may become fatal.

Even if this is not your intention, you must be careful about selecting and processing excessive data that does not give a false impression to the other party.

Graphing is effective for presentations and analysis. While it is possible to create a better graph with various small ideas.

Graphs and visualization skills required for web personnel

Visualizing a list of data is very useful. Web personnel will often create proposals by effectively graphing and expressing the data that supports their claims. On the other hand, in scenes where you analyze access, visualization to read a large amount of data efficiently becomes a very effective tool.

It is essential to make effective use of graphs with the intention of reading and making use of complex data.

In this new series, we will introduce how to select and create graph types so that you can effectively understand what you want to express with data at a glance. Specifically, we will take up four typical graphs (pie chart, line graph, comparison bar graph, scatter graph) and their derivatives, and explain how to use these graphs properly, how to use each effectively, and how to use them dangerously.

How to select and create graphs to process data

Bar graphs, line graphs, pie charts, scattered plots and etc. Visualization of data is indispensable in modern Marketing. But when you try to use the graph function of Excel, you may have trouble using it properly.

In this article, we thoroughly explain the differences and features of the four typical graphs.

Until now, I talked about why graphing skills are important. From now on, I will explain the types of graphs, including their derivative forms.

Each graph will be used properly according to what you want to express, but when you try to make a graph in front of the table, you can immediately judge which graph is best to use. The points of judgment are the three viewpoints of 1) "number of series" 2) "qualitative viewpoint" and 3) "quantitative viewpoint".

Checkpoint 1— Number of series (one or more numbers of columns)

First, look at the "number of series" in the base table. The term number of series is not very important here, but to put it simply, you can think of whether there are one or more columns in the table that are the basis for creating a graph.

When the number is in one column, it is called "one series". If there is only one series, the graphs used will be narrowed down to pie charts, bar charts, and line charts. Below are 2 tables A and B with one series.

Table A: Number of page views of a website from 2005 to 2008. Image by Author

Table B Share of the number of impressions of the banner displayed on a certain web page. Image by Author

Checkpoint 2 – Quantitative perspective (Is the total number 100%?)

If you have one series and the total is 100%, such as the share of sales and sales, use a pie chart. Of the 2 tables above, the total numbers in Table B are 100%, so it is best to make Table B a pie chart.

By the way, you can see at a glance that the total number in Table A exceeds 100.

Figure 1 Pie chart created based on Table B. Image by Author

Checkpoint 3 – Qualitative perspective (Is the title row of the table in chronological order?)

Next, when the title row of the table is in chronological order such as year, month, day of the week, time, for example, in the case of a table that shows the transition of sales every year, you can follow the time series using a line graph. will do so.

Table A corresponds to this in the three tables above. Table A Number of page views of a website from 2017 to 2020

Figure 2 Pie chart created based on Table A. Image by Author

If the title of the table is not in chronological order, for example, if you want to show the sales by product side by side, use a bar graph (vertical or horizontal). As a result, Table C should be a bar chart.

Table C Baner Click Counts on a certain web page. Image by Author

Figure 3 Bar graph created based on Table C. Image by Author

When the number of series is 2 or more

Even if there are two or more series, basically the same check method as one series can be used. I will not exemplify the table and graph, but this time the data before graphing is a table of numbers in multiple rows (or columns).

So for example, sales and profits in chronological order.

If you want to see the transition, use a two-line line graph, otherwise use a comparison bar graph with many bars standing.

This graph is data with multiple series and is used to see the relationship between items or indicators. Specifically, see the figure below.

Figure 4 Example of a scatter plot graph (the relationship between the number of impressions and reach). Image by Author

This is a graph that plots the number of impressions and reach (cumulative arrival rate) of banner ads of a certain medium for each ad material. Each dot represents the number of impressions and reaches of one ad material. For example, an advertising material expressed as a point in the upper right corner has a reach of about 3.2 million people and an impressive number of about 180,000,000.

In general, the more reach you have, the larger your impressions will be, so the points on the graph will line up almost linearly, but sometimes there will be points that deviate significantly from the straight line, such as the points circled in red.

In this example, this point emerges as an advertising menu with the characteristic of having a relatively high reach but a small number of impressions. Scatter plots are useful for understanding the relationships and characteristics between multiple indicators in this way.