Once you’ve cleaned your data, it’s time to start analysing it. Whether you’re conducting market research, assessing your customer experience or running an employee pulse survey, these steps will help you start making sense of your online surveys.
1. Identify your key questions, and the results that feed into them
Before you start survey data analysis, refresh yourself on your basic research goals. For example, your overall goal when collecting data may have been to determine how satisfied people were with a service or product. Questions that speak to this goal might include things like their overall rating out of ten, or how likely they were to recommend to a friend or family member. You could also look at whether they said they’d purchase a second time, or whether they felt it offered value for money.
Grouping the results of a survey together in a goal-focused way like this will help you see patterns beginning to emerge in the data.
2. Cross-tabulate and filter your results
Dig deeper into your survey responses by separating answers out according to the characteristics of the people who answered. For example, you might be interested in whether people aged 35-44 liked your product better than the 45-60 age group, or you might be looking at differences between respondents of different genders.
Cross tabulating means creating individual tables for specific survey questions, including rows with subtotals for your different respondent sub-groups. This type of data presentation is also known as a ‘contingency table or ‘two-way table’. You can express these sub-totals as a percentage, rather than a numerical value, to get a better idea of the proportions of each group in relation to the total.
You can also choose to filter out certain types of respondent, for example those under 18 or living outside the geographical area you’re interested in, then re-tabulate the data to zoom in further on specific results. This can be helpful when you have a wide range of respondents and a wide variety of results. Remember though, you’ll be decreasing your sample size when you do this, which could affect whether your results are still representative or statistically significant.
3. Check your data is representative
As we’ve mentioned, collecting representative data, i.e. the kind which paints an accurate picture of the sample you’re researching, is extremely important. Otherwise, any conclusions you draw may not be accurate or useful.
Making sure you have the right sample size is important here. That’s because having a larger number of people will help to ‘smooth out’ any anomalies or outliers in the data, and reduce the risk that you’ve accidentally picked out all the people who will answer your survey a certain way, while missing out on the full range of opinions and preferences. Having a large number of survey responses to start off with means you’ll have more freedom to explore the data in-depth while still keeping your data representative.
4. Find your averages and benchmark your data
At this stage, you’re in a position to work out the average (mean) number of people who responded a certain way, as well as looking at the mode (the most common answer) and the median (the result that appears in the dead centre of the range). You can also make these calculations for your filtered and cross-tabulated results.
Where things get even more interesting is when you benchmark your data against previous results – also known as a longitudinal analysis. By looking at the outcomes of previous times you ran the same survey as part of a survey program, you can discover changes, improvements and trends in your data over time.
Want to streamline your survey program?
Qualtrics iQ tools give you the option to automate your data analysis, including running statistical tests such as t-tests and regression analyses. Our Stats iQ tool gives you answers on the statistical significance of your findings in plain language – no expert interpretation required. The iQ suite also includes Text iQ, which analyses qualitative data from question types such as free text, using AI and natural language processing to determine overall sentiment towards your brand or product.