- Poor data visualization can lead to incorrect decisions
- Seven data visualization mistakes to avoid
- Avoiding data visualization mistakes is an incremental part of data-driven decision-making
Data visualization is supposed to support fast and valuable decision-making. But just as it may become a part of the path to a breakthrough, it might also hinder a proper assessment of the organization’s state and the ability to find patterns for improvement.
It’s one thing to implement data visualization techniques, and another to truly make use of them. Only when utilized wisely, can they ensure data-driven decisions and put your business on the fast track to success.
We’ve already discussed data visualization practices which speed up time to insights on every organizational level. This time around, *instinctools BI experts will provide you with hands-on guidance on how to avoid data visualization mistakes and implement smart strategies for data visualization.
Poor data visualization can lead to incorrect decisions
The thing with bad data presentation is that you can’t say it’s bad until you fail to get anywhere using it. Such a situation is risky: with all these pie charts scattered all over your reports and tons of descriptive text being classic examples of bad visualization, you may have an illusion that you’ve successfully handled the ever-growing amount of data, whereas, in reality, this data fails to tell the story it’s supposed to. The consequences of bad data visualization can significantly impact your business routine and decision-making processes, here’s how:
- Can’t tell a clear story with your data. Imagine that sitting in on a sales meeting, you expect to see a concise chart with the revenue of the top-five markets where your company is presented, so that you can decide which ones are worth investing more. But you get a pie chart with all the markets where your company is present instead — 20+ slices, most of which are similar in size, making it difficult to evaluate and compare data at a glance.
You may say that at least some kind of visualization is better than no visualization at all. Not really. Unclear visualization doesn’t carry out its functions so you still have to dive into the spreadsheets to make sense of your data.
- Get invalid insights that lead to wrong decisions. What if you’ve tested several new markets to promote the company’s services, and, in one of the regions, the profits for the year increased significantly. But it doesn’t necessarily mean the best match for promotion. What about advertising costs in this market? Maybe they are the highest too. Did they pay off? What’s the return on investment? You should consider the profitability of the advertising costs to understand which market is the most beneficial variant for the organization’s promotion. How many customers reached you thanks to the advertising campaign, and how many through other channels? If the graph doesn’t show the correlation between profits and advertising costs, visualization becomes lopsided and can give you a distorted view of the situation. Taking into account only the revenue numbers, you risk spending a huge budget on advertising in a region with few potential customers.
However, sometimes a couple of graphs are not enough to get a good grasp of a situation. To make a truly well-informed decision, you might need as much as a custom dashboard that helps to track multiple data metrics in one place.
Want to check out the dashboards we built for a recruiting department of a tech company? Get the case study.
- Fail to uncover hidden correlations between different data sets. Weaknesses to improve, anomalies to correct, unobvious interrelationships won’t be revealed with bad visualizations. For organizations, this means the loss of potential revenue and the inability to change the perspective to see new possibilities for development and growth. With misleading data, you won’t be able to define the room for improvement and notice probable pitfalls; for example, you might overlook ineffective marketing campaigns and keep spending your budget on them.
Check our list of common misleading data visualization examples that our BI experts have prepared if your relationship with translating data into images is kind of “continually-trying-to-figure-things-out.”
Seven data visualization mistakes to avoid
It’s unlikely that you’ll make flawless decisions 100% of the time unless you are The Sorting Hat from “Harry Potter”. But it’s possible to minimize the risk of your data being misleading.
#1. Choosing the wrong data visualization technique
There are two tricky moments here. We’ll show them using pie charts as an example:
- Viewers can’t see the difference between slice sizes and, thus, compare them. When the numbers don’t vary much, it’s better to visualize them in a bar chart.
- Viewers can’t get the real dependencies between the objects of correlation. Pie charts are usually used for the comparison of the different parts of a whole. They are suitable for survey results or budget breakdowns (the same pie). But if you use them to compare separate datasets (different pies), you get a bad chart, and data becomes misleading. E.g., a pie chart isn’t a bright idea to compare the number of inhabitants in different areas. It’s better to use a bar plot because human perception primarily judges distances and not areas.
Therefore, choose a visualization technique according to the data’s nature: quantitative data requires charts or histograms, while qualitative information is better presented in pie charts or bar graphs. And make sure the sum of the sectors equals 100% because a pie chart is supposed to show the relationship of each slice to a whole. Otherwise, the viewers will get a mathematical stroke due to your visualization.
#2. Overloading viewers with data
The human brain processes images 6x-600x faster than words. Given the fact that during the next three years, the amount of human-made information is going to triple, the role of good data visualization is only becoming more important. Presenting data in graphics and charts allows us to process huge amounts of information, understand it better, and get insights faster.
But the processing capacity of our conscious mind is still only 120 bits per second. “And what does it have to do with bad graphs?” you may ask. Such a limit for data traffic means that we can’t properly concentrate on the highly-detailed visualizations for a long time. In the case of charts, if there are too many variables, choose 5-6 more essential ones. Graph views with more than 15 items distract attention and may be as frightening as an Excel table with dozens of rows.
#3. Selecting unconventional colors
This mistake usually takes three forms:
- Absolute vs. relative coloring. The function of color is to add extra meaning or dimension. When you use absolute colors, you may miss meaningful nuances.
During the presidential race in the USA, you can see election maps where red is used to mark the states where the Republicans won, and blue indicates the states that supported the Democrats. Such an approach results in maps like the one on the left. It gives an impression of an unquestionable victory of the Republicans, ignoring the fact that in one state people voted for representatives of both parties.
Relative coloring allows the viewers to see a more detailed picture. Looking at the map on the right you can see the proportion of the counties that voted for the Republicans or Democrats against the total number of votes in each county. That way the situation no longer seems so straightforward.
- Unusual colors. Green commonly means something positive, while red is used for negative cases. So if you use them in the reverse way, it may become an example of misleading data visualization. Look at these two maps with pandemic statistics. The first one illustrates the amount of COVID-19 testing in countries all over the world. Due to the green color, even a high amount of cases in China can be taken as a ‘good’ situation. At the same time, white areas don’t indicate the spots where the number of infected is less; white color can simply mean the lack of tests done on these territories.
The opposite example is also true. Look at the COVID maps from the New York Times. You can quickly see hot spots in different American states and figure out the interconnection of the number of cases with the number of vaccinated people.
Also, map charts usually leverage different shades of one color family: the lighter the shade, the smaller the number, and vice-versa. A solution with different colors instead may confuse the viewers. And take into consideration the chance that viewers may be colorblind, so don’t use misleading colors.
- Invisible color on a white/black background. Don’t choose yellow for crucial metrics in a line graph, as it’s easy to miss them on a white screen. The same is true for the picture on the right, where you can’t properly see the borders of the black area on the gray background.
#4. Using uncertain scales
It’s challenging to compare figures with different scales straight away. Inconsistent scale can mislead the viewers. For instance, why is the visualization on the left an example of a bad graph? With it, you aren’t able to assess the scope, it’s not immediately obvious that one figure is four times bigger than another. You should look at the Y-axis and count, whereas good data visualization should exempt you from unnecessary calculations.
#5. Omitting data
Excluding some information, you miss the context. Such an attitude can affect data interpretation. Look at these two graphs: in one case, information is tracked every second year, in another, each year. The left scatter plot is a perfect example of a bad graph because it gives the impression of stable growth, while in reality there are dips and spikes.
#6. Truncating Y-axis
This type of misleading data visualization occurs when the Y-axis doesn’t start from 0. The result of the scale compression is an increasing difference between bars. That way small variations may look paramount.
#7. Operating 3D Graphics in an improper way
3D data visualizations are entertaining and fascinating but the creation of them might not be worth the effort. It’s nearly impossible to follow the height of each bar to the correct Y-value on a multidimensional bar chart. Moreover, you won’t be able to see the values of the bars hidden behind more prominent columns. If these indicators aren’t necessary, just exclude them from the data visualization. If they are crucial, use a simple bar chart instead of a 3D one. Also, remember that setting up all three-dimensional views and colors takes more time and money than making a visualization in 2D.
3D donut charts and pie charts are also more like “hmm” than “hooray” solutions for data visualization. Here is a bad example of a pie chart in 3D. Tilting the pie distorts the image of transparent slices. You can’t define the borders of slices and see how each slice relates to the others and the whole. Additionally, you can’t read labels and figure out which one goes with which slice. Honestly, if Bear Grylls hosted “Running Wild” in the business analysis world, transcribing this pie chart would be in one of the episodes. But you can benefit from the assistance of *instinctools analysts to beat misleading data visualization and create seamless data-filled charts that matter.
Generally, three-dimensional graphics in visualization are rarely used since not everyone has volumetric thinking, not to mention that 3D graphics are as good as being dynamic. Volume is necessary to show an extra axis. Such graphs are used mainly in finance, where bubble charts are needed to show a correlation between funds, stocks, or the stock and the general market. Financial data usually have more than two data series so it’s better to illustrate it with three Axes and different bubble sizes. Since the task is more complex, be especially aware of common data visualization mistakes. Otherwise, you risk ending up with an unreadable bubble chart where it’s difficult to understand if a sphere is larger according to the S-axis or if it seems more prominent because it is closer to the viewers on the Z-axis.
Avoiding data visualization mistakes is an incremental part of data-driven decision-making
Being aware of widespread mistakes can’t level up your business decision-making power all by itself. But just like appropriate usage of data visualization techniques provides you with valuable information and, thus, accelerates your organization’s development, poor data visualization can imperil all that. Bad graphs and charts aren’t the types of challenging situations where you need to fail to be great. When data visualization is misleading, you just make the wrong decisions unknowingly. If your decisions continuously don’t lead to the expected results, check whether you’ve chosen the proper techniques for data presentation. With good data visualization you’ll be able to grasp large volumes of data in no time at all and track changes in real-time. You’ll also get a fresh perspective on facilitating the development of your business — armed with the right data presented in the right way.
Have difficulties with getting actionable insights from your data?
Whether data gives incorrect insights or is just hard to understand, it results in poor business decisions that affect your company’s revenue. Visualizations are bad if they don’t tell a clear story and don’t give the opportunity to uncover unobvious patterns between data sets.
Bad data visualization may be a consequence of insufficient data cleansing because messy information is unlikely to provide you with high-quality analysis. A poorly defined infrastructure and lack of attention to the data preparation stage are the basis of a data-problem iceberg. The tipping point of which is choosing unsuitable data visualization techniques, which lead to poor business outcomes.