top of page
Writer's pictureDimitris Adamidis

Choosing the right data visuals for your results. Get your correct data with the right visual type.

Updated: Mar 17


Data Visuals

In the era of data-driven decision-making, leveraging data is crucial for making informed choices, mitigating risks, and ensuring long-term business success. The Graphiti solution offers a powerful tool to empower your Go-to-Market (GTM) organization, unlocking the full potential of your data assets and driving sustainable growth. By embracing this transformative shift, you can harness the opportunities presented by your data while navigating the associated challenges, ultimately making data your most valuable asset in achieving your business objectives.


Alta - Analytics Platform
 

A study by Tableau found that 75% of data analysts struggle to find the right visualization for their data. Similar to Qlik's study found that 63% of data analysts say they spend more time choosing the right visualization than they should. Finally, Domo found that 57% of data analysts say they have made decisions based on inaccurate data visualizations. Do you still think this is an admin or minor problem?


The definition of proper visual selection is simple. Selecting the right visuals for the correct data is the art and science of choosing the most effective way to communicate data insights to a specific audience.


As always, it's easier said than done. It can't be done in one step without a thinking discipline driven by the data you have selected, insight or context, and finally, the receiver (your audience). Each discussion separately ends badly. The last often forgotten goal of the visualization. That goes back to the previous article, where we underlined the difference between a report and an analysis. The point I'm trying to make is about the key takeaway or call to action you want your audience to make after you present the data to them. It can be random and must be driven by the thinking discipline you apply throughout the analysis process. Otherwise, you end up with a grocery bag that can feed you, but that's not a dish yet, and you still need to cook.


Another ongoing problem with the visuals is the problem calibrating the visuals to your message. We try to cover that in one shot with our selected examples. This time I'll use the examples of charts and focus on their functionality.


In the end, the important criterion for a graph is not simply how fast we can see a result; rather, it is whether, through the use of the graph, we can see something that would have been harder to see otherwise or that could not have been seen at all. Let's keep that in mind while we go over the rest of this article.


There are hundreds of variations, and deciding which one can be challenging. In most cases, it's helpful to consider the different categories of data visualizations and determine which one fits your particular use case. After identifying the right chart type, you can decide which chart best communicates your insight. In the following seven categories, you'll recognize some of the most common charts that business professionals encounter or find useful.

  • Comparison: These charts highlight similarities and differences between discrete values for items or categories. For instance, a bar chart can effectively showcase the revenue performance of various products, allowing you to identify both the top and bottom performers.

  • Trend: Trend charts are utilized to visualize the behavior or performance of a specific metric over time. For example, a line chart can illustrate the monthly fluctuations in inventory levels over the past year, providing insights into inventory trends.

  • Composition: Charts of this nature depict the relative proportions or sizes of different components that constitute a whole. A pie chart, for instance, can demonstrate how a budget is allocated across different expenditure areas, clearly visualizing the distribution.

  • Relationship: Relationship charts showcase the connections and correlations between variables, enabling the identification of outliers, correlations, and clusters. A scatter plot is a suitable example, as it allows you to examine the relationship between customer contract size and satisfaction scores.

  • Distribution: These charts provide insights into how values are distributed across a range and reveal central tendency and shape information. An example is a histogram, which allows you to observe the dispersion of hospital patients by age range, providing a clear understanding of the distribution pattern.

  • Spatial: these charts overlay data onto geographic or spatial maps to uncover behavioral patterns, outliers, or other relevant insights.

  • Flow: depict the movement or transition from one set of values to another through various nodes, connections, or stages. A Sankey diagram, for example, can effectively display the flow of website traffic from the homepage to other pages or sections, allowing for a comprehensive understanding of user behavior.


Let's touch on the other part of the problem around your visuals and message clarity. When you assess whether your message and visuals are properly calibrated, there are three key areas you can focus on:


  1. Keep comparisons close together: Whenever possible, place data elements that are being compared in close proximity to each other. Comparing side-by-side data points is much easier than comparing those at opposite ends of a graph. By ensuring close proximity, you facilitate seamless comparisons for your audience.

  2. Establish a common comparison baseline: When using stacked bars and charts, aligning the values with a shared baseline makes comparisons easier and more accurate. If you want your audience to focus on specific data series, ensure they have a common baseline. For instance, consider using a panel bar chart instead of a stacked bar chart to provide each data series with its baseline, facilitating easier comparisons.

  3. Ensure consistency in charts for effective comparisons: It is vital to maintain visual consistency when presenting multiple charts for comparison. Even subtle inconsistencies can unnecessarily burden your audience, hindering their ability to consume the data and follow your key points effortlessly. To prevent confusion, ensure consistency in aspects such as axis scales, colors, and labeling across the charts being compared. This fosters a seamless comparison experience for your audience.


To get this right, you need to take a step back and ask yourself two basic questions:

  • Does each use case align with an appropriate chart type?

  • If the use case matches its chart category, are you using the most effective data visualization option within that category to communicate your points?

Example 1: Bar Chart vs. Lollipop and Dot Plot Charts

Data Visuals

In certain scenarios, when you have a large number of high values, a lollipop chart may be a better visual option than a bar chart. With a bar chart, the density of the chart will increase due to the length and thickness of the bars. However, because the stems of the lollipops are only thin lines, they can convey the same values with much less ink. But before you abandon bar and column charts for lollipop charts, they have an inherent drawback: each value is found at the center of the lollipop's circle, which is less precise than the straight edge of a bar. These bar chart alternatives are aligned to the same position-based perceptual task that makes bar charts useful for accurate comparisons. However, until they are included as default chart options in more data visualization tools, they won't be nearly as popular as the ubiquitous bar chart.


Here is a hypothetical example. Let's say we want to compare the closing prices of Apple, Microsoft, and Google stocks over the past 5 days. We could create a bar chart to show this data.


The bar chart would show the closing price of each stock on each day, with the bars for each stock stacked on top of each other. This would allow us to easily compare the closing prices of the three stocks across the 15 days (perhaps more accurately relate them).


However, the bar chart would show the closing price of each stock on each day, with the bars for each stock stacked on top of each other. This would allow us to easily compare the closing prices of the three stocks across the 15 days. However, the bar chart would not be as effective at showing the data distribution. For example, it would be difficult to see how many days each stock closed above or below its average closing price.


On the other hand, a lollipop chart would be a better choice for showing the data distribution. The lollipop chart would show the closing price of each stock on each day, with a line connecting the value to a dot. It would make it easier to see the range of values for each stock and the outliers in the data. For example, the lollipop chart would show that Google stock closed below its average closing price on 8 of the 15 days, while Apple stock closed below its average closing price on only 3 days.


The lollipop chart would show that Google stock closed below its average closing price on 3 of the 5 days, while Apple stock closed below its average closing price on only 1 day.


A dot plot would be the best choice for showing the distribution of the data if we wanted to show the closing prices of all the stocks in the market. The dot plot would show a dot for each stock each day, with the dots for each stock stacked on top. This would allow us to easily see the distribution of the closing prices for all the stocks in the market. For example, the dot plot would show that more stocks closed below their average closing price on 8 of the 15 days than those that closed above their average closing price.


Example 2: Dumbbell and tadpole chart as a varIance alternative.


Both of these charts have very similar advantages and disadvantages when it comes to comparing them to simple variance. Please note that the number one mistake for many analysts is mixing up the variance and standard variation (Yes, I'm also surprised). This cardinal mistake is followed by not accounting for outliers, simple mistakes in the variance formula, or implications of variance. Some individuals have variances that can't tell them much about the data distribution. That's a dead end, especially if you run an analysis at the eleventh hour, around midnight before the morning meeting. Yeah… we all have been there.


In general, we need to know that both alternatives offer more robust outliers than just variance, and for the same reason, it's easier to interpret the variance. Comparing the spread of data across different groups is also possible. It's just nice to have that option. On the disadvantage side, it might be a bit harder to calculate than just a variance. In general Tadpole chart is less popular than Dumbbell. I find both very useful while reading The Economist (big fan).


Let's use our previous stock example and apply it to the dumbbell and tadpole charts.


The dumbbell chart would show two vertical lines, one representing the 25th percentile of the stock prices and the other representing the 75th percentile. The area between the two lines would be the interquartile range (IQR).


The tadpole chart would show the same two vertical lines but also include the minimum and maximum values of the stock prices.


If the dumbbell chart is narrow, the stock prices are clustered around the middle of the distribution. This is a sign of a healthy market. If the dumbbell chart is wide, the stock prices are spread out over a wider range. This could be a sign of a volatile market. Well, bingo! You couldn't figure that out while sitting in the room without an expert or using variance only. With that, you solve a second common problem about messaging. If asked to assess the investment in a specific stock, you would reveal a more significant problem around the market volatility. I presume someone in the room is expecting you to know that.


The tadpole chart can help us to identify outliers. If the minimum or maximum value of the stock prices is far outside the IQR, it could be an outlier. Various factors, such as news events or technical glitches, can cause outliers. Why does this matter to the investors? Well, they can skew the results of the measures of spread, such as the std deviation. For example, let's say you have a dataset of stock prices. The mean stock price is $100, and the standard deviation is $20. It means that most stock prices are between $80 and $120. If you don't know the outliers and you trade on that day, it can blip on your screen in red. That can also push the investor to not invest in stocks with many outliers as a sign of volatility.


Conclusion: When selecting chart types to convey your story, your priority should be ensuring your audience can easily follow the information. Clarity, rather than simplicity, should be your goal. Depending on your audience's level of curiosity, data literacy, and patience, you may opt for a less familiar or more complex chart that remains clear. However, it's crucial to consider the perceptual tasks involved and the inherent ease or difficulty of perceiving the data.


As you craft the visuals for your data story, you may encounter the need to make tradeoffs in how you present your data. Depending on which data elements you wish to highlight, you might have to adjust the chart's structure to align it better with your messaging.


While framework models are available to help you navigate the extensive range of chart options, the underlying principles for effective data visualization remain consistent across the board.



12 views0 comments

Komentarze


bottom of page