Statistical Analysis of Customer Survey Data

Customer feedback surveys are everywhere these days. When you buy something online, attend a webinar, or complete a project with a vendor, it's common for them to ask you to complete a survey at the end.


As a marketer, I'm often responsible for creating these surveys, analyzing the results that come in, and drawing conclusions. In this post, I will discuss how to analyze survey data.


Create an Effective Customer Survey

It starts with creating an effective customer survey. You need to be intentional with the survey questions. What are you hoping to accomplish and find out through collecting the answers? Maybe you hosted a webinar and you're trying to gauge what other related topics the attendees would want to hear about next time. Or, you're trying to identify a subset of the attendees to follow up and sell your services.


The specific questions will depend on your goal, but in general, they should be as clear and short as possible - you don't want to waste people's time with unnecessary questions. If the survey is too long, people might give up answering. Try to follow the MECE principle for designing the questions.


Of course, you want to encourage your audience to complete your optional survey. The more responses you get, the more data you have to work with to get clear insights. You can try adding incentives such as being entered in a lottery to win something, sharing a confidential whitepaper, etc.


Once your questions are ready, you can package it up with services such as SurveyMonkey or typeform, which I personally use to create a feedback form for my freelance clients. Typeform has an interactive and modern UX, which should also keep the respondents engaged and encourage people to complete the survey.


Analyze Survey Data in Excel

Excel should be enough for most survey analysis purposes. The type of analysis really depends on your question format (multiple choice, numerical scale, etc.) but you get some initial insights with a few quick bar graphs or pie charts, or mean/median calculations.


You can also use pivot tables and cross-tabulation to segment respondents and make sure that your data is statistically significant. For example, if you host a webinar, some of the attendees will be potential customers (leads) but others might just be some university students who happened to be interested in the topic. In this case, it'd be useful to segment your respondents and focus mostly on the subset of feedback from potential customers.


Start with quantitative data first, because that way it's easier to understand qualitative data later. For example, if I see that the net promoter score has dropped, I know to expect a higher proportion of negative comments.


For more advanced survey data analysis, If you are proficient in the programming language R, it's certainly a powerful tool. It's useful for even qualitative analysis of comments. For example, you can create a word cloud to visually capture the big picture. Another method is sentiment analysis - an automated process based on machine learning, for sorting comments into categories such as positive/negative or neutral.