Chapter 17
Showing Sentiment with Net Promoter Score
Net Promoter Score dashboard.
Author: Steve Wexler
Organization: Data Revelations
Scenario
Big Picture
You are a market researcher and need to track consumer preferences for several brands. You need to be able to see how different respondents feel about each brand and how opinions have changed over time.
Although you will consider asking your survey panel many different questions, you know for sure that you will present the classic Net Promoter Score (NPS) question—“Would you recommend this product or service to a friend or colleague? Please specify a rating from 0 to 10”—as your company has standardized on NPS as a measure of customer sentiment.
Specifics
- You need to show the NPS for a variety of products to see which ones people would recommend and which ones they would not.
- You need to see if people in different roles would recommend or not recommend a product.
- You need to see how NPS has changed over time for a particular product, both overall and for people in different roles.
Related Scenarios
- You need to see results for a series of Likert-scale questions asking survey respondents to indicate the degree to which they agree or disagree with a series of statements.
- You need to see how favorably or unfavorably people rate different candidates running for office.
- You want to know how often survey respondents use different social media outlets and how this usage differs based on age and gender.
Understanding NPS
In an NPS survey, respondents are presented with the question “Using a scale from 0 to 10, would you recommend this product/service to a friend or colleague?”
- Anyone who responds with a 0 through 6 is considered a detractor.
- Anyone who responds with a 7 or 8 is considered a passive (or neutral).
- Anyone who responds with a 9 or 10 is considered a promoter.
The NPS is computed by taking the percentage of people who are promoters, subtracting the percentage of people who are detractors, and multiplying that number by 100. (See Figure 17.1.)
Figure 17.1 How to compute NPS.
How People Use the Dashboard
In its initial state, the dashboard shows only ratings/rankings for several different products. (See Figure 17.2.)
Figure 17.2 Initial state of the dashboard shows only ratings/rankings for the seven products being studied.
After a product is selected, you can see both how people in different job roles feel about the selected product and how NPS for the product has changed over time. (See Figure 17.3.)
Figure 17.3 NPS by role and over time for Product D.
Figure 17.4 What analysts think about Product D over time.
Figure 17.5 Drop-down toggle.
Figure 17.6 Showing percentages instead of NPS.
After a particular role is selected, you can see how NPS for the selected product has changed over time for that role, in this case analysts. (See Figure 17.4.)
Note that at any point, people interacting with the dashboard can change the view to instead show the percentage of promoters, neutrals, and detractors rather than the NPS. (See Figure 17.6.)
Why This Works
The Easy-to-Understand Combination Chart
The main visualization (see Figure 17.2) is a combination chart that combines divergent stacked bar charts with an overall score (the circles).
The divergent stacked bar makes it very easy to see how sentiment skews either positive or negative. (See Figure 17.7.) That is, the entire bar moves either left or right to show which products have a more favorable rating.
Note that, in this case, half of the neutral respondents are on the positive side and half are on the negative side, as we want to show these responses and how they center around zero. (For an alternative approach to dealing with neutrals, see Figure 17.19 later in the chapter.)
The Neutrals Tell Us a Lot
A typical NPS chart shows just the scores, not the distribution of positive, negative, and neutral responses. (See Figure 17.10 later in this chapter.) Neutrals represent the big tipping point with NPS because folks who selected a 7 or 8 are just one point away from being either promoters or detractors. A product with a large percentage of neutrals presents a great opportunity to turn respondents into promoters.
Figure 17.7 Sorted divergent stacked bars make it easy to see which products are more favorably rated.
Breakdown by Role Provides Additional Insight
The ability to select a product and see its performance by a particular demographic (in this case, role) enables users to see how sentiment differs based on that demographic.
In Figure 17.8, we can see that the NPS for Product A among doctors is 35 but is –2 among students.
Avoiding the Traditional Approach to NPS
Consider the snippet of NPS survey data shown in Figure 17.9 with responses about different companies from people in different roles.
Figure 17.8 NPS can differ widely depending on respondent demographics, in this case, role.
If we just focus on the NPS and not the components that comprise it, we can produce an easy-to-sort bar chart like the one shown in Figure 17.10.
Figure 17.9 Raw NPS data about different companies from people with different occupations.
Figure 17.10 The traditional way to show NPS.
Yes, it's easy to see that Company D has a much higher NPS than Company H, but by not showing the individual components, we're missing an important part of the story. In particular, the neutrals/passives are right on the cusp of becoming promoters, so their sentiment is vitally important.
Figure 17.11 Here we see the problem when you just show the results of an aggregation. The Net Promoter Scores for both product X and Y are the same, but they are derived from very different responses.
For example, an NPS of 40 can come from:
- 70 percent promoters and 30 percent detractors or
- 45 percent promoters, 50 percent neutrals, 5 percent detractors.
Same score, big difference in makeup. (See Figure 17.11.)
Figure 17.12 Table with survey results.
Figure 17.13 Likert scale questions using a bar chart.
More Thoughts on Visualizing Sentiment
We looked at why the traditional approach to showing NPS often falls short. What about Likert scale survey data, survey data that asks people to indicate the degree to which they agree or disagree with a series of statements? Let's look at different approaches and see what we should avoid and what we should employ.
Why Default Charts Don't Work Well
Consider Figure 17.12, which shows the results from a fictitious poll on the use of various learning modalities.
It's hard to glean anything meaningful from this figure. What about a bar chart? (See Figure 17.13.)
Wow, that's really bad. What about a 100 percent stacked bar chart instead? (See Figure 17.14.)
Okay, that's better, but it's still pretty bad as the default colors do nothing to help us see tendencies that are adjacent. That is, often and sometimes should have similar colors, as should rarely and never.
Figure 17.14 100 percent stacked bar chart using default colors.
Figure 17.15 100 percent stacked.
So, let's try using better colors. (See Figure 17.15.)
Figure 17.15 is certainly an improvement, but the modalities are listed alphabetically, not by how often they're used. Let's see what happens when we sort the bars. (See Figure 17.16.)
It's taken us several tries, but it's now easier to see which modalities are more popular. But we can do better. Figure 17.17 shows the same data rendered as a divergent stacked bar chart.
Figure 17.16 Sorted 100 percent stacked bar chart with good colors.
Figure 17.17 Sorted divergent stacked bar chart with good colors.
Of course, we can also look take a coarser view and just compare sometimes/often with rarely/never, as shown in Figure 17.18.
Figure 17.18 Divergent stacked bar chart with only two levels of sentiment.
Author Commentary
Steve: I find that the divergent approach speaks to me, and it resonates with my colleagues and clients.
I don't have a problem comparing the magnitude of the neutral percentages that we saw in Figure 17.7, but some of my colleagues suggest that you may want to isolate the neutrals in a separate chart, as shown in Figure 17.19.
Figure 17.19 Alternative approach to dealing with neutrals in a divergent stacked bar chart.
Here we have a common baseline to compare the positives, the negatives, and the neutrals as opposed to Figure 17.7, where the neutrals center at zero.