Learn / Blog / Article

Back to blog

How to analyze survey data and get to know your customers (without any headaches)

If you don’t understand your customers, it’s almost impossible to improve your product or service for them.

Online surveys make it easier than ever to get to know your users, but even a short, simple survey can leave you with a headache-inducing amount of user data to sort through.

So how do you make sense of it all?

User research
UX design and analysis

Last updated

18 Aug 2022

And how do you extract actionable insights that help you improve your product or service for your users?

This article looks at the different methods you can use to analyze and make sense of your survey results. We cover:

The golden rule of surveys: start at the end

Before we get stuck into survey data, let's look at the big picture: while understanding traditional metrics and statistics can be helpful for analyzing survey results, the real key isn’t analysis—it’s planning.

To get meaningful user feedback, you need to ask the right survey questions. And to know what the right survey questions are, you should start at the end and establish clear goals.

Ask yourself:

  1. What’s the one problem we need to solve, or the one question we need to have answered, above all else?

  2. Which customer segments can give us this insight?

  3. When and where should we ask them about it?

  4. What type of survey would best achieve this?

With a clear understanding of the information you need, you can decide which questions to ask and determine how to conduct your survey.

Then, after you've collected actionable survey data from relevant audience segments, your next step is to decide which kind of survey analysis to perform.

💡 Pro tip: choose your analysis methods according to the type of questions you asked and the type of data you gathered. For example:

Closed-ended questions give you quantitative data (i.e. numerical data) that you can analyze with graphs, charts, and comparison tables. Respondents have to answer in a specific way—like giving a 'yes' or 'no' answer, selecting a 1-5 rating, or choosing from predetermined responses.

Use closed-ended questions to measure and compare how your users think or feel about a specific issue. Surveys with closed questions are one of the most effective ways to create customer satisfaction surveys.

Open-ended questions give you qualitative data (i.e. descriptive, word-based data), so you can identify trends and organize responses into groups and categories. Respondents can answer in any way they choose, in their own words.

Use open-ended questions when you want to dig deeper and find out why users think, feel, or behave a certain way.

5 ways to analyze quantitative data

These five methods will help you get valuable insights from quantitative survey data:

1. Make simple comparisons to identify customer preferences


If you asked a multiple-choice question in your survey, the answers will help you identify specific customer preferences (e.g. when you're testing a new product feature, service, or design).

To analyze comparative data, add the total number of responses for each multiple-choice option. Then create a comparison chart to organize the number of responses or percentages for each answer:

This kind of comparison is often easier to understand if you display it in a bar graph or pie chart:


2. Use cross-tabulation charts and graphs to compare results from different audience segments

If you included survey questions that let you categorize respondents, you can see how different audience segments answer the same question—for example, how answers vary by age group or industry.

To analyze these responses, use a cross-tabulation chart to compare answers from each segment:

3. Analyze scale data using mode, mean, and bar charts

If your survey asked users to answer by selecting from a scale, you can measure how strongly customers feel about specific topics, product features, or services you provide. This type of question is also a great way to understand the customer experience.

For example, you might ask respondents to choose from a scale involving numbers:


Or a more descriptive scale:


In either case, one of the easiest methods to analyze scale survey data is to create a bar chart showing the frequency of responses.

Using a spreadsheet like Excel or Google Sheets, enter the names of each response in one row, then add the frequency it was selected in the corresponding row above. You can then create a bar chart that looks something like this:


If your scale involved numerical responses (e.g. a 1-7 rating), there are two other easy ways to analyze the data:

1. Calculate the mode

Mode represents the most common answer that appears in a set of data and can give you a quick snapshot of which rating on the scale respondents chose most often.

You can calculate mode in a spreadsheet with the MODE formula function. Select a blank cell, type =MODE into the formula bar, then select the cells containing the individual responses from survey data.

2. Calculate the mean

The mean is what we generally refer to when we say ‘average’. You can calculate it by adding up all the scores and then dividing the total by the number of responses.

Using the mean gives you a figure representing the typical response, which is helpful if you want to compare how customers’ responses to the same question are changing over time.

Note: there's some debate about using averages to analyze scales, but we'll leave it up to you to decide whether the method will work for your survey analysis.

4. Calculate your Net Promoter Score® using a simple formula

When you ask customers how likely they’d be to recommend your product or service (on a 0-10 scale), you can use the responses to calculate your Net Promoter Score® (NPS).

NPS is represented as a number between -100 and +100 and is considered a practical gauge of customer satisfaction and loyalty. However, it’s typically calculated using a different method:

How to calculate your Net Promoter Score®:

  1. Add up the promoters (those who scored 9 and 10)

  2. Add up the detractors (those with responses 0 to 6)

  3. Calculate the percentage by dividing the number of promoters by the total number of responses

  4. Repeat step 3 for the detractors

  5. Apply the NPS formula: percentage of promoters minus percentage of detractors

Note: you can also find your score using our handy (free) NPS calculator tool. For a more detailed breakdown, read our guide for calculating your NPS.

🔥 If you're using Hotjar

There's no need to go through the process of manually calculating your NPS: Hotjar automatically calculates your Net Promoter Score® and gives you a visual breakdown in your survey response tab.

5. Use benchmarking to compare your data with industry averages or with your previous results

When you’re analyzing quantitative survey data for the first time, it can be hard to tell whether the numbers are as good (or bad) as they seem.

For example, you might learn that your landing page's conversion rate is 12%, which seems low. But if you then learn that the average conversion rate in your industry is 8%, suddenly you’re doing pretty well (yay!).

This is called 'benchmarking', and it’s a great way to interpret your data in a helpful context.

You can learn a lot more from your survey data by benchmarking it with past results.

For example, if you compare your current Net Promoter Score® against your score from six months ago, rather than labeling your score ‘good’ or ‘bad’, you can get a clear picture of progress over time.

🔥 If you’re using Hotjar

When you're viewing survey data in the response tab, filter the results by date to compare answers from different time periods.

Create a product experience your customers will love

Product experience insights from Hotjar help you improve UX and build products that delight your customers.

3 ways to analyze qualitative insights

Qualitative data is often most helpful when used to support and explain quantitative data (more on that later).

But here are three easy ways you can analyze insights from raw qualitative survey data:

1. Create visual representations of survey data


Qualitative data from surveys often involves hundreds or thousands of unique answers. (We feel your pain.)

One great way to process this information is to generate a word cloud from terms that frequently crop up in responses (hint: you can do this with Hotjar). While word clouds won’t necessarily provide a definitive answer to a question, they can be a great jumping-off point for discussion.

2. Read individual responses to uncover hidden insights to shape your product and messaging

Qualitative data isn’t always about spotting trends: it’s also about uncovering the motivations, objections, and desires that your audience won’t (or can't) tell you about directly.

Sarah Doody, Author of UX Notebook, used qualitative data to shape the messaging of her UX training courses:

“A question I like to ask is this: ‘Before taking this course, the biggest challenges I faced were __________,’ or ‘the thing that was holding me back was ___________.’”

When asking questions like these, you’re likely to get many different answers. It can seem overwhelming, but taking the time to read each response can give you a deeper understanding of what your audience really needs from your product.

3. Turn qualitative insights into quantitative data

If you have enough qualitative survey data, you can sort your responses into categories and use them to create graphs, tables, and charts. To learn how to do this, read our guide on analyzing open-ended questions in five steps.

Here are the five steps in short:

  1. Add your responses to a spreadsheet

  2. Look for ways to categorize individual responses

  3. Assign each response to a category

  4. Organize your survey data by categories

  5. Represent your data visually

7 steps to analyze survey results from scratch

1. Establish your analysis goals

Start by deciding the key questions you want to answer with your survey data. This will help you determine which data to analyze, and in which order.

2. Remove any incomplete or unreliable data

If your survey asks multiple questions, you might find that respondents skip some questions or leave fields blank.

This isn't always a problem, but if you’re trying to compare how respondents differ in their answers to multiple questions, incomplete responses may skew your results. In this situation, it’s better to remove the incomplete responses.

🔥 If you're using Hotjar

When viewing your survey data in the responses tab, use filters to remove incomplete answers from your analysis.

Start by looking at the quantitative data that ties most closely to your analysis goals.

For example, imagine you have a SaaS product and you want to understand why more users aren’t signing up for a paid plan after completing a free trial. To find out, you might ask free trial users to rate how useful they find the product.

Now imagine the survey data says 70% of trial users found the product useful. Your conclusion might be that the product isn’t the reason that trial users aren’t continuing.

Looking for trends in this way allows you to hone in on the problem through a process of elimination.

🔥 If you're using Hotjar

Dig deeper and better understand user behavior by comparing data from on-site surveys and session recordings.

Place an on-site survey on the page (or pages) you want to investigate, analyze the survey data, then watch session recordings of users on the same page(s) to see how they experience and interact with your site.

Note: you can also test this in reverse order! First watch session recordings of the page(s) in question to identify blockers or UX issues, then use on-site surveys to ask direct questions and discover ways to improve UX and the customer experience.

4. Check that your findings are statistically significant

Drawing meaningful conclusions from quantitative survey data can be tricky. Data often suffers from ‘noise’, because people sometimes make mistakes when entering their answers (but not you!).

If you only have a handful of responses, that ‘noise’ will affect survey results even more. So the less data you have, the less likely it is that your findings will be statistically valid.

Use our sample size calculator to check that your data pool is big enough to trust the insights you’re finding, and make sure you’re not jumping to conclusions in your data by assuming that correlation means causation (more on this later).

5. Compare your quantitative data against previous benchmarks

Whenever possible, try to get a frame of reference for interpreting your data. Looking at historical data can help you to make sense of trends you have identified.

Let’s go back to the example of a SaaS company trying to understand why free trial users aren’t signing up to a paid plan:

Your company might compare its results against benchmarks from a similar survey the previous year.


Now let’s imagine that trial users find the product more useful in the current year than last year, but paid sign-ups still haven't increased. This could indicate that you need to focus on optimizing other factors—like customer experience (CX) or pricing—instead of further developing the product.

6. Use qualitative data to support and explain your quantitative data findings

Using quantitative and qualitative survey data together can help you build a complete picture of what’s happening—and of what customers need from your product.

Where quantitative data often reveals audience trends and preferences, qualitative data can reveal the why behind them.

Let’s apply this to the same SaaS company that wants to determine why free trial users aren't becoming paid users:

By asking a closed-ended question, you identified that 70% of trial users found the product useful, which indicates that the problem probably isn't with the product itself.

If you also asked an open-ended question like “what’s stopping you from signing up to our paid plan?” you could sift through the answers to discover that trial users also objected to the pricing.

🔥 If you're using Hotjar

Responses to survey questions aren’t the only type of quantitative data to look at: use heatmaps and session replays to discover whether users encounter issues like confusing navigation, website bugs, or broken elements.

7. Present your findings to colleagues

Now that you’ve gathered actionable insights from your survey data, it’s time to share your findings with your team.

If you’re sharing your findings in a meeting, remember that it can be challenging for people to digest raw numbers quickly. It’s best to present your findings concisely with charts, graphs, or infographics in these situations.

However, if you’re creating a more detailed report that your colleagues can read in their own time, you can include some more in-depth breakdowns of figures.

🔥 If you're using Hotjar

Bring the voice of your customer to your teams in Slack: when you create a survey, just toggle on Slack under 'Forward Response', and select which channel(s) you want to receive your survey responses in.

Creating a helpful narrative

Whether you’re putting together a visual presentation or a detailed report, try to create a narrative around your survey analysis. Take your team members on the same journey you experienced to reach your conclusions (but feel free to skip any analyses that weren’t helpful or conclusive).

Start by stating your goals, and what prompted you to conduct a survey in the first place. Then, present your datasets in an order that supports your analyses.

In the example of the SaaS company, you:

  • Identified a key trend with quantitative data

  • Used historical benchmarking to eliminate one possible explanation for the trend

  • Looked at qualitative survey data to identify pricing as another possible explanation

  • Harnessed other data (from heatmaps and session recordings) to connect the dots and further verify your findings

How to avoid jumping to the wrong conclusions

A common mistake when analyzing surveys is jumping to conclusions about what causes a particular trend. Statistically-savvy readers (hi there!) often refer to this error as ‘confusing correlation with causation’.

For example, imagine you're gathering data on why restaurants go out of business. Your data might reveal that, surprisingly, every restaurant that failed used cutlery.

You have two pieces of data here:

  1. The number of restaurants failing

  2. The number of restaurants that use cutlery

When you put these two datasets together in a graph, it shows that 100% of restaurants that failed used cutlery— so the datasets seem to have a relationship (i.e. they are correlated).

But it wouldn't make sense to argue that having cutlery was causing the restaurants to fail.

So even though the two data pieces are correlated, it doesn’t mean one is causing the other.

Avoiding the 'correlation and causation' error in your analysis

Make sure you’re not jumping to conclusions with your analyses. Take every factor into consideration that could influence trends, and use qualitative data to uncover the factors you hadn’t thought of. Also, challenge assumptions by looking for data that disproves them.

For example, imagine your benchmarking shows

  • Customer satisfaction with your product has gone down since the previous year

  • Your company updated your product’s features in the previous year so you conclude the updates caused the downturn in customer satisfaction.

There's the error.

Even though the two factors—customer satisfaction and product features—changed at the same time, it doesn’t mean one was causing the other. It could be that customer satisfaction was affected by customer support issues, website loading speed, or any number of different factors (which may have had nothing to do with your product updates).

Use web and behavior analytics tools to investigate and understand what's happening on your website to avoid acting on assumptions—and find out the real reason for the drop in customer satisfaction.

Final thoughts

Analyzing survey data helps you understand customer behavior and track how your company is performing, but working with large amounts of data can quickly become overwhelming.

Keep things simple, and remember to:

  • Design your surveys with clear, simple goals from the beginning

  • Use quantitative data to spot initial trends, then use qualitative data to look for more in-depth explanations

  • Make sure your conclusions are valid by using benchmarks, checking that your sample size is big enough, and considering correlation vs. causation

Create a product experience your customers will love

Product experience insights from Hotjar help you improve UX and build products that delight your customers.

FAQs about analyzing survey data