Visual tricks to find insights in dispersion

Visual techniques and different perspectives were crucial to analyze distribution and identify key consumer drivers for a client in the car industry, and to find the path for a major brand transformation aligned with the expectation economy.

Even those who know how important it can be to analyze data dispersion may find themselves staring clueless at their charts, which makes it useful to review visual tricks and techniques that can help find insights in distribution. The potential impact of being able to understand data distribution should be celebrated: comprehending how income is unevenly distributed throughout a population is the first step towards fighting social inequality. This article will exemplify distribution with a marketing business case, but it aims to inspire data analysis in other fields as well, because the techniques to explore the data can be applied to many scenarios.

Case: Consumer drivers in the car industry

To create truly relevant data visualization, we should focus on the business questions. One of the marketing challenges of car manufacturers is finding out what drives consumers to buy and to keep engaged with their particular brand.

While working with a big player of the automotive industry, I came across a proprietary survey from dentsu, which ranked different factors that made people feel more connected to different brands throughout the consumer journey: from consideration to purchase to recommendation. Each one of the factors or reasons, that we called “consumer drivers” (unfortunately easy in this industry to confuse with car drivers, the person who drives the car), received two scores based on the survey:

1. Performance = How consumers perceive that attribute in a particular brand. This is calculated as the mean agreement from consumers with each attribute.
2. Derived importance = How important that attribute is to customers in each stage of their customer journey. This is calculated as the relationship between attribute agreement and the customer journey stage KPI.

Let’s go through the steps taken to start the visual exploration of this data and to make it more digestible to both analysts and stakeholders.

Starting the visual exploration

1. Establishing the relationship

As I often say, it’s not enough to have data available, data is only valuable when it’s understood. The first step to begin the visual exploration is defining what we want the chart to help us understand. In this case, it was the relationship between performance and derived importance. One of the best ways to represent visually the relationship between any two variables is to plot them in a scatterplot.

The first version of the scatterplot, created in Excel. It can benefit from a few tweaks to help the interpreter understand more in just a glance.

2. Grouping according to follow-up actions

Data is only as valuable as the decisions it enables. So it’s crucial to define what actions can be taken based on the data and the parameters used to analyze it.

In this case, we used the median value of each score to split the axes, zoning the scatterplot in four quadrants:

1. Strengths: high importance, high performance
2. Improve: high importance, low performance
3. Maintain: low importance, high performance
4. Monitor: low importance, low performance

The second version has increased value as a chart with the addition of actions per quadrant. It could still have more clarity about the attributes and more focus on what is most relevant for the business.

Luckily, the main issues with the second version were easy to fix. With the aid of rectangle shapes at the back of the chart, the colors highlight the areas most important to the business: the strengths (in green) and the attributes that need improvement (in red).

The attribute names in the labels also made the scenario much more tangible. To add labels like that on a scatterplot in Excel or PowerPoint may not be so intuitive: right-click on the labels > select “Format data labels” > on the right side menu, go to the columns icon > Label Options > Value from cells > Select the range in the data source where the attribute or category names are.

The third version made it much easier to understand the data, but we still had to get to the core insights.

3. Exploring alternative perspectives

The dispersed dots on a scatterplot resemble fishes in a pond, and the analyst is an avid and hungry fisher. Some insights may feel like throwing a net and grabbing several fishes, when we’re able to find clusters and to group data points in meaningful ways, like what was done with the quadrants. And other insights need more patience, throwing the bait, staring at the chart until you almost go mad, and holding the fishing pole tight as a single big fish seems to fight your grasp.

When going for the big fish insights, we must explore different perspectives to find the “a-ha” moment, the great catch. And data Visualization is intrinsically linked to analytical thinking when examining a dataset.

In the car industry case, we were dealing with data from four different countries that were relevant to the client brand. That meant four different versions of the scatterplot. But naturally, one of the ways to find differences and commonalities between consumers of those countries was to plot all countries in the same chart.

The fourth version of the scatterplot focused only on the quadrants that were most relevant for the business: strengths and improvements. It has colored dots to represent the different countries, making comparisons possible. However, multiplying the number of dots by four made reading the labels an unpleasant task.

The overlapping labels may be a known challenge for the seasoned analyst, but they also know it’s unpresentable. Our scatterplot needed to be simplified. After diving into that data with a team of media strategists, we dug up 3 strengths and 3 attributes to improve, that were relevant to the client brand and fairly similarly positioned in all countries.

The fifth version showed the core findings, and it may seem elegant from a distance. However, it over-corrected the busy previous version and ended up over simplifying the conclusions. It had no visible evidence of the serious data analysis backing it up.

Making the scatterplot digestible and nuanced

Our last challenges were to bring back the countries’ nuances and the complexity of the data behind the chart, while still making it easy to communicate the insights.

4. Grouping relevant factors

We created three pairs of attributes of relatively high importance, combining one where the client brand was performing well and the other where it tended to underperform:

1. The brand was considered to have a great driving PERFORMANCE, but it was not a brand consumers would DREAM of having.
2. Consumers thought the brand had a great overall DESIGN, but that it didn’t create delightful EXPERIENCES consistently.
3. It was a brand that consumers TRUSTED, but not a brand they ADMIRED.

The consequence of grouping those attributes was the “a-ha” moment: our client’s brand was performing well in “functionality” and underperforming in “feelings”. But in the new expectation economy, consumers demand more from brands. The big catch was to propose a major brand transformation, leaping from “functional product” to a car that benefited consumers’ “feelings”.

The slide that summarized the key insight and the 3 pairs of attributes.

5. Refining the data visualization

Here are the techniques used to refine the last version of the scatterplots:

1. To highlight in the vertical axis what was most important to the consumers, the axes were swapped.
2. To hint at the data on the background without stealing too much attention, the axes labels and gridlines were brought back, but grayed out.
3. To make clearer charts, avoiding labels overlapping and the charts becoming too busy, the labels on each dot were switched to the single word that summarized each attribute.
4. To highlight the desired quadrant, a very light green rectangle shape was positioned in the “strenghts” (high importance, high performance).
5. To easily identify the countries and avoid an unnecessary legend, the dots were turned to flags. It’s fairly easy to do that on Excel or PowerPoint, by selecting each dot > right click > Format data point > Bucket menu > Marker > Fill > Picture or texture fill > Insert > select your image.

The sixth and last version of the scatterplots helped communicate clearly and fast to the client team what were the key insights about their consumers’ drivers and what needed to change in their brand strategy.

Conclusion

The visual techniques applied in the example case, to explore the dispersion of the data and to make the scatter plot more digestible, helped find a path to a much needed change in brand strategy, and to communicate that argument clearly to stakeholders. Similar cases of exploring datasets to find the relationship between variables and their dispersion can benefit from the same steps and techniques.