A guide to survey statistics and analysis
With the right information and tools, you can use statistical methods to analyze your survey data without being an expert.
Once you’ve crafted your survey questions, sent surveys out to your target audience, and collected then responses, you’ve probably got a ton of data on your hands. Now, you could skim through it, choose one noteworthy statistic, and move forward with it. Or, you could do an in-depth survey analysis to find the rest of the gems in your data.
Concerned because you’re new to survey statistics? Don’t worry. We’re here to help shed light on survey statistics and analysis.
Why choose SurveyMonkey?
With more than 17M global users, we’re a leader in survey software. Explore our plans and features to see what works for you.
Survey statistics and statistical analysis
Statistical analysis of your survey results will reveal deep insights into the data. Some of the discoveries may include:
- Whether the trends you’ve observed at a high level are actually meaningful
- What your data means when put into context with other data you have collected
- Whether there is one identifiable factor that is affecting your business more than others
- Where you should direct your next research efforts
- How you can use the insights from the survey data to create meaningful change
Your statistical analysis will help you summarize a large amount of data, and in many instances, allow you to make inferences from your sample to the larger population from which your sample is drawn. Learn more about our survey platform and basic statistics.
Related reading: The risks and disadvantages of data silos
Importance of data analysis
Data analysis makes studying data and identifying actionable insights easier. It makes it simpler to study trends and patterns, such as things that might otherwise have been overlooked.
Data analysis provides:
- Better targeting to reach your customers effectively and impactfully
- Improved insight into your target market
- The ability to predict future trends—and drive innovation
- Informed decision-making for internal budgeting
- Help to solve problems with data-backed decisions
But before you can begin your analysis, you need to conduct survey data cleaning. Cleaning survey data includes identifying and removing any answers from respondents who don’t match your target market or didn’t answer questions thoughtfully. If you skip this step, your ability to capture valuable insights is limited, and your findings' credibility is reduced.
Also helpful before analysis is another instrument in your statistical toolkit—benchmarking. SurveyMonkey Benchmarks is a simple way to compare your survey results with thousands of other organizations. Benchmarking uses weighting to adjust variables that might affect overall results. This information provides you with a “standard” reference to help you identify variances in your data.
Tip: SurveyMonkey has three ways to prep your survey data for easier analysis and reporting.
Survey statistical analysis methods
When you’re ready to analyze your survey data, you’ll want to choose a method that best suits your data and research goals.
Choosing your method
There are several methods for statistical analysis of survey data. The decision of which method to use depends on the level of measurement and the number of variables involved.
Level of measurement
There are four levels of measurement that determine how survey questions should be measured and what statistical analysis method should be used.
Nominal data
Nominal data classifies data that doesn’t have quantitative value. Any numeric scores assigned to response categories are arbitrary. For example, “Choose your preferred toothpaste brand from the list below.” From this data, you can only track how many respondents chose each option and which one was selected most.
Ordinal scales
Ordinal scales classify data that does have a quantitative value that’s used to show the ranking order of the data. For example, ordinal scales that place data in ranks could include: support-oppose, agree-disagree, or excellent-poor rating scales. You can determine median and mode from this type of scale. Ordinal scale data can also be analyzed through cross-tabulation.
Interval scales
Interval scales show both the order and difference between values. It’s a quantitative measurement scale that shows order, a meaningful and equal difference between variables, and the presence of zero is arbitrary. Examples of interval scales for surveys would be age in years or monthly spend in dollars. There is also a quantitative value, and you can analyze median, mode, and mean.
Number of variables
- Univariate analysis consists of only one variable. This is the simplest analysis because only one quantity changes. Its main purpose is to describe the data. An example of this might be a set of heights. Height is the only variable, so you can find mean, median, mode, range, minimum, maximum, and so on.
- Bivariate analysis involves two variables. This type of data deals with the relationship (correlation or association) between two variables. An example would be an examination of the relationship between outdoor temperature and ice cream sales.
- Multivariate analysis involves three or more variables. This is similar to bivariate data. An example would be the effect of education on income, controlling for gender.
As a reminder, the mean is what most of us refer to as the average of a set of numbers, the median is the middle number in a set of values, the mode is the most common number in a dataset, and the range is the difference between the largest and smallest number in the dataset.
Dependent and independent variables
A dependent variable is a variable that is being tested and measured. An independent variable is a component of the research that the researcher can manipulate or change. This independent variable is assumed to have a direct effect on the dependent variable.
As you’ll see, the number and type of variables and level of measurement factor heavily into your decision when choosing a survey statistical analysis method.
Frequency distributions
A frequency distribution is a representation of a survey dataset within a table. It is used to organize and summarize data. It is basically a list of values that a variable takes in a dataset and the number of times each value occurs.
Works best for:
- Levels of measurement: nominal, ordinal
- Number of variables: univariate
- Data display: tables, bar graphs, pie charts, histograms
- Example: Our survey participants were asked the question: How many pets do you have at home? The results were: 3, 0, 1, 4, 4, 1, 2, 0, 2, 2, 0, 2, 0, 1, 3, 1, 2, 1, 1, 3. Our frequency distribution table might look like this:
| Number of Pets | Frequency |
| 0 | 4 |
| 1 | 6 |
| 2 | 5 |
| 3 | 3 |
| 4 | 2 |
T-test
This statistical test is used to compare the mean of two groups or the difference between one group’s mean and a standard value (benchmark). This is generally used when the datasets come from the same population and may have unknown variances. In this case, population can be described as the full set of individuals who could potentially participate in your research and variance as a measure of the range of the responses. A T-test is used as a hypothesis testing tool and to understand if the differences in groups are statistically significant.
Because of this, it allows the following assumptions of the data:
- The scale of measurement follows an interval or ordinal scale
- The data is collected from randomly selected units of the population and is representative of the total population
Tip: While T-tests can tell you if something is significantly different, you will have to determine whether the identified difference is meaningful to your study.
Works best for:
- Levels of measurement: Independent variable is dichotomous and the dependent variable is assumed to be interval
- Number of variables: bivariate
- Example: Do Millennials spend more at our store than Gen Z shoppers? A t-test will compare the spending habits and reveal statistically significant differences.
Analysis of variance (ANOVA) test
There are two types of ANOVA tests:
- One-way ANOVA compares the means of one independent variable with two or more groups to determine whether there is evidence that their population means are different. If there is a statistically significant result, the two populations are unequal (different).
- Two-way ANOVA extends the one-way ANOVA to examine the influence of two different independent variables on one continuous dependent variable
Works best for:
- Levels of measurement: nominal or ordinal independent/dependent variables
- Number of variables: bivariate or multivariate
- Data Display: table, bar graph
- Example: Do people spend different amounts depending on which credit card they use?
Crosstab analysis
This type of analysis uses data tables to display the results of each respondent. It enables you to examine relationships that may not be overtly apparent when looking at survey responses. Crosstabs are used for categorical data—values that are mutually exclusive to each other.
Works best for:
- Levels of measurement: nominal or ordinal independent/dependent variables
- Number of variables: bivariate
- Data Display: tables, clustered bar chart
- Example: Will customers buy my cat perfume? Is there a relationship between gender and intent to buy? In the table above, crosstabs show that while 45% of respondents would buy my product, of that percentage, women are more likely to make the purchase than men (67% versus 33%). Without using cross-tabulation, you may not have discovered that your target audience should be women.
Related reading: Cross tabulation analysis: definition and examples
Regression analysis
In regression analysis, a set of statistical methods is used in the estimation of the relationships between a dependent variable with one or more independent variables. Regression analysis identifies the precise impact of a change in the independent variable.
Works best for:
- Levels of measurement: interval (linear regression) or nominal (logistic regression) dependent variables
- Number of variables: bivariate or multivariate
- Example: A conference host can use regression analysis to understand what factors most impact attendees’ satisfaction
Cluster analysis
Cluster analysis groups data in a way that a particular set of data elements are more similar to each other than those in other groups. There is no dependent variable when clustering, so this method will often indicate hidden patterns in the data. This can also provide additional context to the dataset.
Works best for:
- Levels of measurement: interval
- Number of variables: multivariate
- Example: Find out what features of a cell phone plan are important to smartphone users by having them rate the importance of each feature. Use the data to uncover the underlying personas of smartphone users.
Factor analysis
This method, also called dimension reduction, is a way to reduce the complexity of your findings by trading a large number of initial variables for a smaller number of underlying variables. With factor analysis, you’ll uncover hidden factors that explain variances in your findings. Factor analysis can be used as a pre-step in segmentation.
Works best for:
- Levels of measurement: interval
- Number of variables: multivariate
- Example: Simplify large blocks of data from Matrix Likert Scale questions to focus and clarify results with factor analysis. A Matrix Likert Scale question assigns weights to each answer choice to calculate a weighted average for each answer. You can assign any weight to the answers, including the standard Likert 1-5 scale.
It’s time to analyze your survey data
Statistical analysis of your survey data can seem daunting, but it’s well worth it. You’ll uncover information that can’t be seen in a basic review of your survey results. There are several methods to glean the most relevant insights from your surveys, and if you need help analyzing your data, check out these five best integrations to use with SurveyMonkey.
When you’re ready to make the most of your survey data, start with SurveyMonkey.
Discover more resources
Toolkits directory
Discover our toolkits, designed to help you leverage feedback in your role or industry.
How to make your data collection more secure
Boost survey response rates with these 20 email templates
Enhance your survey response rates with 20 free email templates. Engage your audience and gather valuable insights with these customizable options!
P-value calculator: How to calculate p-value
Leverage our p-value calculator to find your p-value. Plus, learn how to calculate p-value and how to interpret p-values with our step-by-step guide.