Use the Compare tool to contrast two or more groups to find out their important differences. A ‘group’ is simply a named subset of your data you define through a set of filter conditions.
For example, when analyzing Airline Reviews you could have a group for Business Class and one for Economy Class. The Compare Tool will show that there are differences in importance of opinions relating to the Seating, Food, or some other quality between these two groups.
By default the Compare tool will display differences based on 'Contribution' to your target metric. Contribution is effectively 'Impact' - with an adjustment to normalize the group sample sizes so that the comparison is as if each group was 100% if the data.
Two important ways to use Compare
There two key ways to use the Compare tool depending on what you are looking to understand.
Most Important Differences (Most Different) - When comparing groups to understand which patterns have the most important difference - you will likely want to view the patterns with the largest absolute difference in frequency, target metric or both (contribution).
Most Extreme Differences (% Most Different) - The second way to use the Compare tool, is to understand which patterns have the largest relative difference. Using the Compare tool in this mode can often uncover patterns that may have a low frequency, but offer unique untapped insight. For example when analyzing Airline Reviews only 2% of customers might mention 'privacy'. However when comparing Business Class to Economy you may discover that Privacy is rarely mentioned by Economy customers, and that Business Class customers are 23 times more likely to mention it (4.7% vs 0.2%). So understanding this pattern may represent opportunities for improving Business Class traveller experience.
Using the Compare tool (Walkthrough)
Defining groups to compare
The first step in using the Compare tool is to define the groups you wish to compare
Click '+' to add a new group (you can add up to five groups)
When only one group is added it will be automatically compared with its logical opposite e.g. if you add a comparison of the airline 'United Airlines' it will be compared with 'Not United Airlines'.
You can assign each group a custom name. If you leave the Group Name field blank it will be automatically assigned a name based on your filter selection
After defining a group, you can change its definition by clicking on the group chip. Or alternatively delete it and add a new one.
Some interesting types of groups to compare are:
Different customer segments or demographic groups
High Scores vs Low Scores (typically these will be compared by Frequency)
A groups that mentions a specific theme or concept - to everyone that doesn't
Two or more different periods in time e.g. 2017 to 2018
Understanding different aspects of the data
When analyzing data using the Compare tool you can change the 'By understanding' control to view the data through different dimensions.
The most important three views are:
Concepts - Provides an overview of important drivers relating to common concepts
Themes - Provides a more nuanced view than concepts, including aspects unique to your data that may not be included in Concepts
Segments - Highlights important drivers and differences in your structured data. Either select 'Segments' to see all these categorical attributes at once, or select a specific attribute.
Use the Priority control to highlight what is important
Use the Priority control to specify
The metric you wish to compare
Whether you want to see patterns with the largest aboslute difference (Most Different) or relative difference (% Most Different)
The four most commonly used priority metric options used with Compare are:
Contribution - Use when you want to compare groups of different sample sizes, to understand the relative importance in contribution of both the pattern frequency and target metric to each group (this is the default option).
Frequency % - Use when you want to understand relative differences in the frequency of concepts, themes or segment patterns. This is also the best option to use initially with '% Most Different' to find extreme differences between group aspects.
Target Metric - Use when you want to understand the differences in patterns for your target metric
Impact - Use when you want to understand the differences in pattern impact on your overall target metric across your groups (note this is likely only useful when comparing two groups of similar sample size e.g. comparing two years or two teams
For some Datasets when you select the '% Most Different' priority option with Frequency % or your target metric - you will possibly see many low frequency patterns. You choose to filter these out by manually adjusting the Frequency % Cutoff setting in the Priority control to a higher cutoff (say 2-3%).
Visualizing the results
The results are visualized in a Bar Chart where each group is assigned a different color.
Before trying to understand the bar chart, quickly review the legend to familiarize yourself with the key metrics for each of the groups
Note also that you can
Toggle on/off bars for any of the comparison groups by clicking on the group name
Toggle on/off comparisons for 'All Data'
Change the charted metric (for the same patterns) by clicking on the metric label in the chart (Overall Rating in above example).
Also note that when you select a priority metric of NPS, Net, Average or Subset % the chart will display the Overall bars for each group at the top of the chart - as well as vertical line indicating where the mean for the entire data set sits - providing visual indicators of how each pattern compares to its group mean as well as the mean for the current data selection.
Understanding and explaining each pattern
Next compare each of the 'Most Different' patterns - which appear in order of greatest range (difference in largest priority metric between the group with the highest value and the group with the lowest value).
For example we can see that Waiting Time is a pattern that has a negative contribution to both Business and Economy groups - however it has a significantly larger negative contribution for Economy Class (-0.41 vs -0.06).
To better understand how this pattern differs for these two groups mouse over the two bars
Here we can see that
Economy Class customers mention Waiting Time roughly 5% points more frequently that Business Customers (33.5% vs 28.4%).
Although this is a relatively large increase (approximately 20% lift), there is another more important factor here - the Average Score for Economy Class for this pattern is a full point lower than Business Class (5.2 vs 6.2).
The end result is that the negative contribution of this pattern is 7 times larger for Economy than Business Class (-0.41 vs 0.06).
Perhaps you are thinking this major difference in importance partially explains why airlines typically include Business Class ticket benefits that reduce waiting time or help mitigate its effect. e.g. Priority Check-in, Priority Boarding, Chauffeur Service and Lounge Access
To validate this we can click on each bar to explain the two different groups for this pattern. The below results quickly validate this insight and tell two very different travel experience stories.