Techniques for Better Survey AnalysisReading time: 5 minutes
So, you conducted a survey, and the responses are in front of you. You know there is valuable information inside your results, waiting to be drawn out. But where do you start your survey analysis?
While every survey is different, there are a number of tried-and-true practices and procedures to keep in mind anytime you’re evaluating survey results.
In this post, we’ll cover considerations you should make before, during, and after your analysis to ensure you are getting the most insight possible from your survey responses.
Before Diving Into Analysis
First, review and prepare your data. This means compiling the results in an organized format, but it also means cleaning the data to make improvements like replacing null values and dealing with outliers.
Also, at this stage, consider bringing in additional data sources (such as customer information from a CRM or institutional data from a database) if this data adds to an overall understanding of the surveyed population, or to create benchmarks to measure against. Survey data is often structured differently than traditional database tables, so it may require some reformatting to match your tables up.
One key consideration to make before starting your analysis is to determine if there are any new variables you can create from your existing dataset. For example, do you have a zip code for each respondent? If so, it’s fairly easy to determine the respondent’s physical distance from your location. That may be a more valuable data point than mere physical location. Consider if there are any such transformations to make to your data before starting analysis.
Finally, it’s critical to take some time to review the goals of your project and ensure that you are building your analysis and results-sharing procedures in a way that will help you clearly answer the right questions. Filtering, binning, and creating subgroups allows you to focus your results on directly answering those key questions. Creating a data dictionary that explains each of your survey’s codes (for example: on a scale from 1 to 5, is 1 best, or is 5?) is a worthwhile step to take for your end-users as well.
Perform Your Analysis
There are two major types of data you are likely to encounter in your survey results:
- Quantitative Data (values stored as numbers)
- Qualitative Data (open-ended responses)
By nature, each will be treated differently during your survey analysis.
Quantitative Survey Analysis
Generally speaking, it’s best to start your analysis with quantitative data. As Hubspot puts it, “…quantitative data can help you better understand your qualitative data. For example, if 60% of customers say they’re unhappy with your product, you can focus your attention on negative reviews about user experience.” An understanding of some of the hard numbers drawn from quantitative analysis can steer you toward or away from focal points in the more open-ended qualitative results.
The Pell Institute provides a useful, comprehensive guide to quantitative analysis which is worth bookmarking as a reference when working with quantitative data. The guide provides advice on analyzing nominal, ordinal, interval, and ratio data, as well as a walkthrough of tabulation, disaggregation, and other important steps.
Predictive modeling opens your quantitative survey data to a wide range of uses. Predictive analysis allows you to take your static data and make it dynamic. You can game out a variety of scenarios to see what the impact of making changes to your strategy might be, without having to invest resources and time. This is where combining data from your CRM or database can really accelerate your analysis.
Qualitative Survey Analysis
Qualitative response analysis tends to be more complicated than quantitative analysis. The range of answers a respondent can provide is infinite. However, with a few techniques in hand, qualitative data can be effectively and efficiently analyzed toward the end goal of actionable results.
One option is to turn your qualitative responses into quantitative data. To do so, code responses based on common criteria. For example, a question might ask a participant to describe what they enjoyed about a film. You could create a coding system that corresponds to commonly-mentioned features in responess. If a response mentions CGI, you could code that as a 1; cinematography, a 2; and so on. This may involve some interpretation on your part. However, it can help you get an understanding of patterns and trends to view alongside the detail provided in the text responses.
Using a text analysis algorithm makes parsing text for patterns an easy process. This can be used for sentiment analysis: a breakdown of how people feel about a topic. Many of the major digital survey platforms feature a text analysis tool, such as SurveyMonkey and CheckMarket.
Visualization is a third option. A commonly used method of visualizing trends in text responses is a word cloud. These quickly make clear the most prominent words in text. A word cloud gives you a pulse check on what to investigate further in your data. It can also help you determine which values are likely to be worth coding. Note: when visualizing, be sure to remove stop words (“the”, “and”, “or”, etc.) first.
Share Your Findings
Once you have actionable information to share with your organization, it’s time to decide how best to do so.
Consider several factors:
- What level of data is best for your audience: high-level overviews or detailed statistics?
- Will the results best serve the audience as raw numbers or in charts and graphs?
- Is there value in the reports showing live data and updating automatically, or does a static report work best?
- Should the audience be able to create their own dashboards or should they all see the same information?
- What is the technological capability of your target audience?
Reporting offers you the ability to deliver the same information to your audience on an ad hoc or scheduled basis. This is a good path when everyone needs to be on the same page, working from the same information.
Visualizations, with applications like Tableau, allow you to create impressive dashboards with a huge range of design and functionality options. Depending on your own familiarity with dashboard design and the tech ability of the intended audience, you may run into “information overload,” so that is worth bearing in mind.
Custom dashboards with a simple interface are a great step toward data democratization: that is, opening live data up to stakeholders so that they can make decisions based on the exact data they need to see. Rapid Insight’s Bridge enables anyone to quickly and easily build and view custom dashboards on any device.
The majority of the advice in this post was drawn from a recent webinar hosted by James Cousins of Rapid Insight and Will Patch of Niche. The presentation is overview of the full process of gathering, cleansing, and analyzing survey data. View the on-demand video here.
And if you’re looking for tools to make survey analysis more efficient and effective, visit us at Rapid Insight. Our data tools enable you to cleanse your survey data, easily build predictive models, and share insights with your entire organization via custom dashboards accessible from any device.
If you have any survey analysis techniques of your own to share, leave a comment below!