Deriving Insights From Feedback

clear glass bulb on human palm

Today I’m pleased to share a post that was co-written with Dan Heilbron, Head of Customer Experience Advisory Services, Asia-Pacific and Japan at Qualtrics.

A relatively simple way of deriving insight from customer feedback is through correlation analysis.

Correlation analysis is a statistical technique that aims to establish whether a pair of variables could be related. It is used for spotting patterns within datasets. From a CX Manager’s perspective, the technique can be incredibly helpful to understand which independent variables (such as satisfaction driver questions in a voice of customer survey) could be influencing a dependant variable (such as a specific key metric like an NPS score or an operational metric like call hold times).

It is important, however, to note that while correlation analysis will help you to determine whether a statistical relationship exists between an independent variable and a dependant variable, and the strength of this relationship, it will not prove causation – i.e. that one event caused another event to occur. It will also not tell you the exact impact of the independent variable on the dependant variable.

To illustrate this point, let’s consider the CX Manager of an airport who discovers a correlation between the number of people slipping over in the airport foyer and the number of complaints about cancelled flights. Intuitively you’d say there’s no causal relationship between the two and you’d be right. What’s missing is a third variable which explains both: the presence of rain. Water on peoples’ shoes causes them to slip over in the foyer. Rainclouds also cause flights to be cancelled which leads to more complaints from customers.

When conducting correlation analysis, the strength of the relationship between the two variables being analysed is measured using a correlation coefficient (usually displayed as r) which will range between -1 and 1. A positive correlation means that both variables increase in relation to each other, while a negative correlation means that as one variable decreases, the other increases. A correlation of 0 means that there is no relationship between the two variables at all. The closer to 1 or -1 r is, the stronger the relationship between the two variables.

The Advantages of Correlation Analysis

  • It is low cost. The analysis can be done using standard business tools such as Microsoft Excel which are almost universally available.
  • It is relatively simple to understand. It’s not difficult for the average CX professional to learn the principles that apply to correlation analysis and then interpret the results.
  • It provides meaningful insights. The technique provides information about the strength of the relationship between variables and insight into the degree of certainty of those relationships which assists decision making and action planning.

The Disadvantages of Correlation Analysis

As with any statistical analysis technique, there are some downsides to correlation analysis:

  • It does not prove causation by itself. If those doing the analysis don’t apply correlation correctly, there is a risk that variables will be identified as causing one another as opposed to just being correlated to one another. This could lead to CX Managers focussing on the wrong influencing factors of experience. To illustrate this point, let’s look at per capita cheese consumption in the US and the number of people who died by becoming tangled in their bedsheets:

Correlation Pic

An r value of .947 is very high but of course in this case, it’s pure coincidence. The variables are completely unrelated; one does not cause the other. More “spurious correlations” like this can be found here.

It needs to start with a pre-defined set of variables. Managers need to select the variables that they will be testing prior to beginning the analysis. It will not identify the variable(s) that may be the real issue if they’re not part of the initial set.

Let’s illustrate how it’s done using an example. This time the manager of a contact centre wants to study the relationship between Average Hold Time (the independent variable) and NPS scores (the dependant variable) to ascertain if the time a customer spends on hold is related to their NPS score.

Here’s some sample data:

Average Hold Time (secs)

NPS score

































In this example r = -0.92777 – a strong negative correlation showing that when hold times went up, NPS scores went down. In other words, it is highly likely that the longer a customer spends on hold, the worse their perception of the experience would be and the less likely they would be to recommend the organisation.

Tips for Getting the Most Out of Correlation Analysis

  • Ideally run your analysis using a statistically significant amount of data. If you aren’t sure on how big your sample size needs to be, here’s a handy calculator to help you.
  • Use actionable independent variables: If you are running your analysis using key drivers and key metrics from your voice of customer survey, you will want to ensure your key driver questions are as actionable as possible. This will allow you to determine not just which aspects of the experience are having the biggest impact on your CX and operational or financial metrics, but also what action to take to improve those experiences.
  • Consider using qualitative and quantitative data: Some analytical tools allow you to use unstructured data topic sentiment as your independent variables. This gives you the ability to, for example, use respondent comments to see the impact that certain topics mentioned by customers have on CX or operational metrics.


Correlation analysis is not just the domain of people with degrees in statistics. It is a useful tool that all CX Managers have at their disposal to help them draw insights from feedback data. Its relatively easy to understand and doesn’t need expensive software to run. Just make sure you’re studying actionable variables and be careful not to jump to conclusions when looking at your results!

About My Co-Author

Heilbron_Daniel_col_hi (1)

Dan Heilbron, CCXP leads Qualtrics‘ APJ CX advisory team and is based in Melbourne, Australia. Dan spends his time helping clients across the APJ region to maximise value from the strategic, cultural and VOC program design elements of their XM program, and to ensure their program evolves as their business goals and customer needs evolve. Prior to joining Qualtrics, Dan worked with other large XM SaaS organizations in Australia, the UK and the USA and has provided XM advisory services to Fortune 100 and exchange listed clients, government and not-for-profit entities across 5 continents, in industries such as sport and entertainment, automotive, airlines, hospitality, retail, B2B, financial services and healthcare.

Image courtesy of UnSplash.


view all posts

Subscribe to Cxpert

* indicates required

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>