How to analyse survey data

The results are back from your online surveys. Now it’s time to analyse the survey data and turn data into meaningful information you can act on. This guide follows the full path: clean your survey data, summarise the numbers, make connections from open-ended feedback and build a clear story for stakeholders.

In this guide, you’ll discover what survey analysis is, the different types of survey data you can interact with and six steps to conducting effective survey data analysis.

Survey analysis is the process of transforming raw survey data into insights you can act on. It focuses on the patterns, trends and relationships in your responses, using descriptive methods (such as counts, percentages and averages) and inferential methods (such as correlations and significance tests) to interpret survey results and inform decisions.

Good survey data analysis starts with intention because most surveys blend closed-ended questions with open-ended feedback. Going into survey analysis with a plan – clear goals, a small set of priority questions and a sense of who needs the findings – makes it easier to decide how to analyse survey results and which outputs matter most.

Related reading: How to use a survey design and data analysis plan to drive better insights

The type of questions that you ask in a survey will determine the type of data that you receive. Data can either be quantitative or qualitative. 

Quantitative data are numerical responses from closed-ended questions. For example, in a Net Promoter Score® (NPS®) survey, ratings are on a scale from 0–10, and the numbers can be counted, averaged and compared across groups or over time.

Sample survey question: “How likely are you to recommend us to a friend or colleague? (0–10)”

Numerical data is fantastic for tracking progress over time. But if you want to gain insight into why someone responded the way they did, a follow-up question that yields qualitative data is a great idea.

Qualitative data are open-ended responses in which people explain their choices in their own words. These answers reveal motivations, pain points and context behind scores.

Sample survey question: “What is the main reason for your rating?”

When responses start pouring in, it can be hard to know where to begin. These six steps turn survey data analysis into a repeatable workflow. They help you decide what matters most, structure your survey data and build a story that is accurate, fair and easy for stakeholders to act on.

Before starting with survey analysis, you must ensure your survey data is usable. Cleaning your survey data means deciding which respondents to exclude from your analysis, removing low-quality responses and eliminating duplicate responses.

Some survey-takers may only have answered a small selection of your total questions, while others may have rushed through your survey without giving proper consideration to their answers. All of these cases would contribute to low-quality data, which may have an impact on the accuracy of your final results.

Conducting a response quality analysis will help you identify survey responses that don’t match your criteria. For example, if you’re collecting data related to customer sentiment, a response from someone who isn’t a customer wouldn’t be beneficial.

Preparing your survey data for analysis by making it easily readable, clearly formatted and logically sorted will give you a clean and clear dataset to conduct your analysis.

A top survey research question is your guiding force for a survey. These questions relate to your survey research objectives, which you should have determined when setting a survey goal.

Before diving into the granular details of your data, you should start with your top survey question. For example, if your question was “Would people be interested in buying this new product?”, then the following table of results is where you should start:

Would you purchase this product if it were available today?

As you can see, the vast majority of people responded positively. With that in mind, you can proceed with data from other questions to gather more information about this product.

Broader, top-level survey questions will help you locate primary opinions and ideas within your data. Once you’ve laid this foundation, you can move into more intricate parts of data analysis to respond to your research questions. 

Often, one of the best ways to gain a better understanding of data is to break it down. Your business can use crosstabulations (crosstabs) to divide your data into smaller groups based on shared characteristics. For example, you could filter the data according to survey-takers from a specific background. You can filter, compare and create rules to analyse data about each use segment.

Data filters are a powerful way to gain comprehensive insight into your data. For example, if you wanted to know how a product fared with different groups, you could include demographic questions in your survey. Since people have to specify their age group, you can instantly filter responses from each group. Comparing these datasets would then show how much each age group likes your product, helping you instantly spot outliers or less-happy demographics.

You could filter by:

Although qualitative data doesn’t yield the percentages or raw numbers that you can use here, you can always use text analysis tools. By gathering written responses and putting them through one of these tools, you could create a sentiment analysis that shows how your customers feel. 

Converting raw data into insights involves proving statistical significance. After all, you must ensure that your data shows a finding rather than just cherry-picking cases. Statistical analysis reveals whether the trends that you observe are meaningful and what the data suggests when used with other datasets. 

There are several methods you can use to prove statistical significance:

Beyond raw analysis, other factors determine whether or not you have meaningful results. For example, the total number of responses and completion rates compared to your total sample size will suggest whether or not your results represent a statistically significant group.

Equally, it’s a good idea to consider the margin of error in your survey results. A margin of error reflects how closely your results mirror the entire population. A small margin of error suggests that you have high confidence in the accuracy of your results, whereas the opposite suggests that your results could fail to represent the wider world. 

If you need more survey responses to add a level of authority to your survey results, SurveyMonkey Audience has you covered. 

Benchmarking is the process of comparing your results to those of other companies or industry averages. You can even benchmark your results against yourself with repeated survey experiments that collect data over different quarters or years. 

Businesses around the globe use customer experience benchmarking to determine how their customers perceive them compared to their competitors. 

By conducting a longitudinal analysis, you’ll discover how your survey results have changed over time. For example, if you host an annual conference, you could ask attendees to rate their satisfaction with the event by using a conference feedback survey. If the average satisfaction rate decreases over time, you have an interesting entry point for further research.

You could couple the satisfaction question with an open-ended one, asking people to explain their choice. This qualitative data will shed light on what may be causing the decreasing satisfaction, helping you take actionable steps to fix it. 

You can even track data for different subgroups. If, for example, satisfaction rates are increasing year on year for marketing staff but not for administrators, you might want to look at administrators’ responses to various questions to see whether you can understand why they are less satisfied than other attendees.

Your survey data has the power to tell a story. Start with your top research question, outline your findings and go from there. What did you want to find out? What did you learn from your data? What specific findings stand out or are particularly interesting?

Where possible, use visuals to support your report. Your readers won’t want to plough through paragraph after paragraph of written numbers. A simple graph or word could summarise your results and enable people to intuitively understand what the data is saying. 

Now that you’ve analysed your data and found statistical significance, it’s time to show it to the world. Creating a great report showcasing your findings is the final step of practical survey analysis. With this report, you can share your findings with colleagues, which can inspire change in your organisation. 

Related reading: How to use six basic charts to create effective reports

Even careful teams run into the same pitfalls in survey analysis. These usually fall into two categories: problems that weaken the findings and reporting habits that make results seem stronger or narrower than they really are. Recognising these patterns makes it easier to spot them early and keep survey data analysis grounded in reality.

Use these patterns as a quick check before you finalise your survey data analysis, so common mistakes don’t slip into your findings or reporting.

The most prominent mistake people new to research make when conducting survey analysis is confusing correlation with causation. Causation occurs when one factor directly causes another. Correlation, on the other hand, occurs when two factors move together but may not be related.

A great example of this is the effects of cold weather. Drinking hot chocolate and wearing mittens are correlated variables, as they both tend to go up and down together. However, one does not cause the other. They are both caused by a third factor: the cold weather.

Just because two factors move together, that doesn’t mean they are related. Performing correlation analysis will help you identify the relationships between your data.

You could connect two pieces of data with a tertiary motive if you confuse correlation and causation. 

As we mentioned, you can think of your survey results as telling a story. However, you could be missing the bigger picture when you cherry-pick data, that is, selecting key data points that validate your hypothesis.

When you set out to prove something with a survey, it can be frustrating when your data doesn’t tally. Perhaps some customers simply don’t agree with a key statement, or your data doesn’t show a statistically significant majority. Whatever you encounter, the last thing you should be doing is being hyper-selective about the data points you use.

If you’re having to pick and choose which data to use in your calculations, then you’re undermining the accuracy of your survey results.

SurveyMonkey knows better than anyone that conducting surveys and gathering results can be extremely exciting. That said, when you rush to calculate survey results, you could be overlooking important data that hasn’t yet been filled in. 

Surveys often need large pools of responses to increase your level of confidence in the data. If you rush to calculate results as soon as possible, you may have fewer responses than you need in order to prove statistical significance. 

Imagine if your survey results show that 100% of customers are interested in a new product feature. Although that sounds amazing, if only one person responded to your survey, how confident can you be that your pool of responses really reflects your wider customer base?

The questions that you include in your survey directly correspond to the type of data and the quality of the data that you receive. With that in mind, your survey data is only as good as your survey questions. 

Here are a few tips:

By enhancing the quality of the surveys that you write and send out, you’ll naturally begin to collect higher-quality data.

There are many aspects of survey data analysis, from looking at your top results to slicing and dicing the data to reporting on the results. 

SurveyMonkey has survey analysis tools and templates that’ll make it easier than ever to reach your target audience and get reliable responses. 

Sign up to use templates and tools created by experts.

NPS, Net Promoter and Net Promoter Score are registered trademarks of Satmetrix Systems, Inc., Bain & Company and Fred Reichheld.