The best insights you can get for your business are obtained through customer surveys. They go beyond your sales revenue and goals and look deeper into what your customers like and dislike about your products, services, and brand as a whole.
The act of collecting customer satisfaction survey results is the first step. Next, you need to understand how to analyze, summarize, and present this survey data to provide actionable insights you can use to make changes to scale and grow your business.
In this step-by-step guide, you’ll learn how to analyze survey data, including our 8-step process for performing a detailed survey data analysis:
1. Start with your end goals
2. Conduct your survey
3. Remove incomplete data
4. Check if data is statistically significant
5. Analyze and evaluate data
6. Compare data to benchmarks or historical data
7. Explain results with qualitative data
8. Summarize and present findings
1. Start with your end goal
Unlike a good book, always start at the end when planning a satisfaction survey. This will help ensure that when you get to the feedback analysis stage, you’re likely to have the data you need.
Your survey goals could include:
- To better understand your brand recognition and trust in the marketplace
- To see if newly launched products or features are useful
- To plan your product lifecycle based on understanding user needs and wants
- To create a benchmark for future business growth
Once you have your goal, you can write questions that relate to that goal.
Step 2: Conduct your survey
How you write and distribute your survey will impact the data you have to analyze when it’s done:
Decide your data collection method
There are many types of surveys you can use to collect your data. Your choice will depend on your end goal (which you created in step 1) and the available resources to analyze collected data. If you use a machine learning method for survey analysis, you have more options, but if you plan to do manual survey analysis, you’ll want to choose your data collection method carefully:
- Interview-style: You have a standard list of survey questions verbally given to the respondents. You have someone writing down or recording their answers through video or audio.
- Focus groups and observations: Focus groups work well for observing how people interact with a product in addition to answering specific questions about it. With focus groups, the respondents’ answers could be influenced by hearing the responses of others, so keep that in mind when analyzing your data. Regardless, the qualitative observational data it provides can be quite valuable if that relates to your survey goals.
- Written survey: You can also send paper or online surveys. A benefit of online surveys is that your data is already entered into a spreadsheet or downloadable data set, so you’ll save time inputting data into a computer for data analysis.
Determine sample size
Any feedback is good, but if you’re surveying to understand the larger picture, you want your sample respondent size to represent the whole. For example, if you have 100 customers but only survey 5 of them, their responses may not be an accurate representation of how the majority of customers feel.
You can do complex calculations (such as Slovin’s formula) to determine your exact sample size, but an industry average is usually a maximum of 10% of your population.
Then, take an educated guess to determine how likely people are to do your survey. Do you have a vocal, engaged group willing to provide feedback and expect most people to respond? Or do you feel uptake will be lower at a 10% response rate? Take that and send out enough surveys, so you will likely get your 10% response rate.
When you have designed your survey and determined how many people you need to send it to, it’s time to begin distributing or administering your survey. For the highest response rate and most accurate responses, ensure it’s clearly written, and there is no ambiguity in the questions.
Online surveys are ideal because the data doesn’t need manual digitization. Your survey questions and answers are already in the computer. It’s much easier to manage online surveys rather than paper ones when you can.
Unless it’s an ongoing survey (like one that’s sent with customers’ digital purchase receipts), have a clear deadline for collecting responses.
Step 3: Remove incomplete data
Once you’ve collected all your data, it’s time to put it into a format to make it easy to do your survey analysis. Often this means two parts:
- Entering quantitative data into a spreadsheet
- Coding qualitative data so it’s more easily summarized and interpreted.
Then, review your data and determine if any incomplete or irrelevant data can be excluded from your final survey report.
For example: Let’s say you send a survey to 100 people. You ask people if they feel the price of your product is fair with a yes/no response. If 50 people leave that question blank (unanswered), but 40 people say “yes,” and 10 say “no,” you can’t say 40% (of 100 people) answered yes. Instead, it’s actually 40 of 50 people who responded positively, which is 80%. In this case, a non-response doesn’t automatically indicate either of the options so it’s considered incomplete data and shouldn’t be considered in your totals or calculation as either “yes” or “no.”
When sharing this data, you should distinguish that 80% (of 50 respondents) responded this way. Or, you must clarify that 10% said no, 40% said yes, and 50% were undecided or non-responsive.
TIP: If you’ve put this data into a spreadsheet, you can filter out non-responses by writing the following formula into your cell:
“First cell” and “last cell” represent the range of data you want to check. The result will calculate the total number of empty cells (aka non-responses), so you can subtract them from your sums and averages for that question.
Step 4: Check if data is statistically significant
This step may be optional, depending on the goals of your survey. We know that any data is helpful, whether it’s statistically significant or not; however, if you want to get the general sentiment or opinion of your entire audience, ensure you’ve collected enough data so it’s representative of the whole.
If it’s not, you can still get some valuable data, especially if you asked qualitative questions.
Step 5: Analyze and evaluate data
When you’re confident you have enough information to satisfy your goal, it’s time to analyze your survey data.
How do you evaluate a survey?
At its core, there are two ways to analyze survey data: Manually or by using algorithms and machine learning. Both will use the following types of data analysis on your survey results.
What statistical analysis should I use for questionnaires?
Your choice of data analysis methods will depend on the data you have collected and what you hope to learn from the data. Here are two popular methods of analyzing your statistical survey data:
Regression analysis: This type of statistical analysis looks at the relationship between two or more variables in your survey data. It looks at how one variable affects (or influences) another variable.
For example, you may wish to know what factors contributed to someone being satisfied (or dissatisfied with their purchase of your product:
- By looking at other dependent variables, you may see that those who were happy with the purchase were more likely to note that they loved the extra product support they received.
- By looking at those who were unsatisfied with the product, respondents also often mentioned that they liked its packaging but said using the device operation was confusing.
This is a great way to start developing a narrative around your data and learn how to use it to inform business decisions based on how your customers (or survey respondents) feel.
Longitudinal analysis: When analyzing survey data from a single data collection source (like a survey), you get a good picture of your current state. And, examining the same data over a longer time helps you identify trends and growth (or recession). This is called a longitudinal analysis: how responses to specific questions change.
To do this ethically, you must use the same questions in more than one survey distribution. This works excellently for surveys after an annual event or conference. It can help you see which events were perceived by attendees as successful. If you do quarterly surveys to measure customers’ satisfaction with your product, you may find that people enjoy your product more in the summer than in the winter.
With this information, you can better understand trends in your niche and use that information to uncover the reasons behind those ebbs and flows. When you know what influences these trends, you can use your survey data analysis to inform business decisions to help predict them and prepare accordingly.
Survey analysis for quantitative data
Analyzing survey data that contains quantitative data (numerically based data) is often quicker and easier to understand and draw conclusions. There are several ways to collect this numerical data for survey analysis, including:
The nominal scale: Respondents choose the best response from a list of given labels. These questions ask people to “choose from the following,” then follow with a list of unrelated items for the respondent. The responses can be limited to one answer per question or can allow for multiple answers. You can then use the data to discover which the most popular answers were, based on how many people selected that option:
|What was your favorite topic in this course:
When analyzing this data you simply add up each response to get your totals.
Ordinal scale asks people to rank their value or agreement with a given statement or question: Usually, these are ranked on a scale of 1-10 but could be 1-5 or a similar ranked sequence of responses:
|How strongly do you feel about the following statement: The staff served me in a timely manner:
Strongly Agree | Agree | Undecided | Disagree | Strongly Disagree
When analyzing this type of survey data, you can calculate the percentage of people who responded to each answer.
Interval scales are questions in which the responses include a scale with equal distances (known as intervals) between each value. There is no true zero in this type of question. A basic example of an interval scale question is:
|What is your highest level of education obtained
As you can see in the above example, each of these represents the different levels of education, in order. Based on the responses to this question, you can calculate the average education level of your respondents.
Ratio scales are similar to interval scales but have a true zero. It’s a form of quantitative data. Here is an example:
|What’s your age:
You can see that each age group is the same (9 years).
Which methods are used to analyze survey results?
When analyzing survey results, you can also do three other simple calculations to analyze the data:
- Mode: Mode determines the most common answer from your survey respondents. You can calculate the mode of any numerical dataset in Excel by typing =MODE() and then select the cells that contain the data you are trying to summarize.
- Mean: Mean is the statistical way to say average. This mean value will give you the “typical” response to a question. You can calculate the mean average of any numerical dataset in Excel by typing =Average() and then select the cells that contain the data you are trying to summarize.
- NPS: If you are doing a Net Promoter Score (NPS) survey, you can calculate your Net Promoter Score (NPS) from your data. To calculate your NPS score, you use the following formula.
Survey analysis for qualitative data
Analyzing qualitative (non-numerical) data is much different and often more time-consuming than quantitative data. When reviewing qualitative survey data, there are two ways to get valuable insights from this data:
You can measure customer satisfaction using sentiments. This means drawing conclusions about your respondents’ answers and expressing their sentiments. It can be as simple as a positive or negative sentiment or multiple levels of like or dislike. It’s about categorizing the answers to open-ended questions into themes (in this case, emotional responses).
For example, you can interpret the response “I love your product” as positive sentiment and “The price was too high” as a negative sentiment.
From this, you turn the response to open-ended questions into an answer that can be analyzed numerically (i.e., the percentage of people who are happy with the product).
Coding qualitative data
Coding qualitative data means assigning categories or values to each written or observed response. These values can then be added and averaged to determine an accurate overall representation of each area of your business you are analyzing.
Once you have all your qualitative data quantified, you can calculate and begin drawing conclusions about the data.
Turn non-binary responses into binary responses
Often your survey will ask yes/no questions or questions where there are only two options. To assist with your data analysis, do a find-and-replace in your spreadsheet to turn these options into binary responses (1s and 0s).
For example, if you ask, “Did you buy this product online?” in your spreadsheet, turn all “yes” responses to a 1 and all “no” answers to 0. With this data, you can easily calculate the SUM of that column to indicate how many people said yes to that question.
Step 6: Compare data to benchmarks or historical data
Once you’ve got your survey data in a readily accessible format, it’s time to compare survey results to industry benchmarks or historical data from your organization. You can skip this step if you don’t have access to any benchmarks. With this information, you can get a snapshot of your progress in certain areas of your business.
For example, for the past two years, you’ve sent every customer a brief customer satisfaction survey with their purchase receipt. One year ago you presented a big report to senior management about where people were dissatisfied with your product offering. Now, you can take the data from last year and compare it side-by-side with your data from two years ago to see if your scores have improved. This can also point out areas of decline.
If your company doesn’t have comparable data to analyze your current survey data against, you can look for any industry benchmarks available. For example, your financial services company subscribes to an industry association that did its own study and found that 25-30% of all customers in the industry generally become repeat customers. With this data in hand, you can compare it to your returning customer score to see how you compare.
How do you present survey data?
When presenting current survey data with past survey data, be sure to note which are the recent scores. Then you can visually illustrate your findings via graphs, charts, or narrative.
How do you tabulate survey results?
Using cross-tabulation and filters, you can better analyze and compare your results. For example, if you’ve sent out a survey to three key audiences, and you want to compare how they answered each question, you can tabulate your responses to make it easier to see your results:
Q: Would you attend this community event again?
|Yes||No||Undecided or Non-responsive|
If you analyzed individual questions, 51% of all respondents said they would attend this event again. Looking at this result, it appears that the event was not very successful if your goal was for people to return to future events; however, diving into the data deeper, you can see that 80% of moms loved the event. This means that if you cater your next event more towards moms, you could expect higher overall average satisfaction ratings next time.
Correlation vs. causation
You can also use data analysis to look at the relationship or influence between survey questions to see if they are correlations or causations.
A correlation is when two variables move together. They do not influence each other but move up or down in similar paths. They are caused by a third factor which may or may not be something your survey will uncover.
A simple example of this would be sales. When you sell more mittens, you usually sell more scarves. This is influenced by the season (not something your survey will bring to your attention as it wasn’t a survey question). The result is knowing that your business typically sells more of these products when the weather gets colder (the third variable)
Causation is when one variable affects the other. For example, your data may show that when your sales of alcohol go up, guest satisfaction also goes up. If you had no other data points, you could assume that the more people drink, the more they enjoy their time at your event.
Step 7: Explain results with qualitative data
Going beyond just the numerical data, you can use your qualitative data to help explain your results in your survey report. This enables you to build a narrative (story) around the data to better understand what it’s telling you to get more actionable information.
Knowing people are satisfied with your product is a good first metric to learn. The next step is to use qualitative data to find out what they liked about it and what they don’t like, if it exists.
Create a narrative
Explaining your results by telling a story can be a helpful way to present your findings, and qualitative data can help you build this story.
A comprehensive survey data analysis can help you create narratives better to understand areas of your business, including:
- Your target demographics or customer personas
- Your marketing funnel
- Your customer journey
Step 8: Summarize and present findings
Many people find that writing a survey report of your findings is easiest by creating a PowerPoint deck with numbers and stats. We recommend a mix of data and insights (aka what the data means).
A basic example is the data may tell you that 75% of customers don’t plan to renew their subscription when it expires. While this is an important stat, also share what that means: 75% of customers don’t intend to continue because of your recently announced price increase.
When summarizing your findings, try not to jump to conclusions. Look for multiple pieces of information to back up your assumptions before calling them a fact. For example, if 75% of people don’t plan to renew their subscription, and you also see a decrease in annual household income levels, it’s not necessarily a safe bet that household income is the reason (or the only reason) for the non-subscription renewals. If you acted on this assumption, you might miss vital data and the actual causation.
The actual reason people are not subscribing is more likely linked to the fact that your help desk has reported an increase in customer service inquiries in the past six months. Based on survey data, your customers are unhappy about your upcoming price increase. It’s more likely that these two factors are contributing to the low renewal rates. Finding the right causation between your data points is critical to know how to act.
Finding the right relationships and making the most accurate conclusions can be easier when you use machine learning and AI to analyze data from surveys and other connected systems (like helpdesk support tickets).
How do you summarize survey results?
Once you have the data you want to share with your team, prepare a summarized version of the results and the raw data. This will help the “number nerds” who like the numbers and those who want the actionable insights or narrative.
Here are a few ways to summarize your survey results:
Visuals (Charts and graphs): To present one to three survey questions, show them in a chart or graph. Avoid stuffing too much data into your charts, and always summarize the key findings from each chart.
Infographics: Infographics are great for those who process data best when it’s presented visually. Infographics also make for great sharing on social media if that’s something you’re planning to do with your results.
Compare with benchmarks and historical data: Create ways to compare your current survey data with either industry benchmarks, data from competitors, or your historical data. Use this to share what’s changed and predict future trends and performance.
Sentiment analysis: Knowing how your customers feel about your brand is critical. You can summarize your findings by customer sentiment to understand who your happy customers are and what makes them happy.
Use storytelling: Consider weaving your data and analysis into a narrative when sharing your results. Here’s an example:
“Our survey found that before customers bought XYZ Software, they spent too much time and energy manually analyzing their survey results. Most (80%) were referred to our company by someone already a customer.
Once becoming a customer, these referred customers recorded fewer support ticket requests than other new customers, likely because they could troubleshoot any problems and get their answer from the person who referred them. However, nearly 92% of customers noted having issues reading the UI due to the color combination we’ve used. We predict that if we change the font color from yellow to a dark gray/black, we can increase customer satisfaction on the App.”
Using machine learning for survey data analysis
Manually analyzing your survey data can provide some great insights, but you can get more actionable insights using machine learning and AI. With a platform like Idiomatic helping you process and analyze survey data you:
- Spend less employee time doing manual survey analysis.
- Pull data from other systems (like attendance and registration systems, helpdesk systems, and customer databases) to look for additional correlations or causations in your data.
- Quickly analyze open-ended feedback into actionable insights
Idiomatic is an AI-driven customer intelligence platform that can help you make sense of your survey data by generating a survey report with real-time actionable insights. You get analyst-quality insights without waiting days or weeks for staff members to comb through your data. We help you unlock the “why” in your customer feedback from survey responses.