{"id":15680,"date":"2021-06-28T13:00:00","date_gmt":"2021-06-28T20:00:00","guid":{"rendered":"https:\/\/devwww.3cloudsolutions.com\/post\/techniques-to-make-your-survey-data-sing-3\/"},"modified":"2024-06-18T09:13:12","modified_gmt":"2024-06-18T16:13:12","slug":"techniques-to-make-your-survey-data-sing","status":"publish","type":"post","link":"https:\/\/3cloudsolutions.com\/resources\/techniques-to-make-your-survey-data-sing\/","title":{"rendered":"Techniques to Make your Survey Data Sing"},"content":{"rendered":"

With the ubiquity of online survey services like SurveyMonkey<\/a><\/span>, Qualtrics<\/span><\/a>, and many others, it\u2019s very simple today to get a survey out to a large group of people and collect responses in a timely manner. The question is, what can you do with all that data that comes back to get the most value out of it? Recently, I did some volunteer work for a local government agency to produce visualizations for print, using Power BI<\/a><\/span>, and I want to share some techniques that worked well. If you are an advanced Excel user, and perhaps are wondering what benefit there is in newer tools like Power BI and Power Query, this post is for you.<\/p>\n

<\/p>\n

\"Blog_Singing
\n<\/span><\/p>\n

Problem<\/span><\/h2>\n

Traditional survey results from services can be a bit lackluster because the services are general-purpose, and they don\u2019t \u201cknow\u201d your data intimately. The appearance of charts and graphs or typeface and color selections may be limited and, most important, getting deeper insights into your results may require extra steps and extra time. This is especially true of doing analysis of the data across<\/em> different questions.<\/p>\n

<\/h2>\n

Suggestions<\/span><\/h2>\n

Get the Data
\n<\/span><\/span><\/h3>\n

The first step is to pull the results out of the service in a basic, raw data format such as an Excel spreadsheet or CSV file. Examine what you\u2019re getting back in terms of format \u2013 what are the columns? How are different types of questions represented? Are they text, numerical values? For a given question, are the responses spread across different columns?<\/p>\n

<\/span><\/p>\n

We aren\u2019t going to look at survey design issues in this post, focusing on how to consume the data. But one basic technique is to ask some demographics questions in addition to questions about the topic at hand. For each survey question, make a note of which category it belongs in \u2013 is it about the respondent? Or about the topic? The demographics answers are data we want to transform into dimensions – that is, data that describe all the other questions.<\/p>\n

Another way to think about this is the \u201cby\u201d phrase \u2013 what are the questions in your survey that you want to see \u201cby\u201d other responses? For example, is it favorite color by age group?<\/em> Whatever is after the \u201cby\u201d is the dimension you will want to apply across<\/em> all the topic-related questions.<\/p>\n

Next, examine the questions by question type. Setting aside the topic content altogether, note which questions are numeric or text responses, which are multiple choice, which are ranking and so on. It\u2019s likely that you\u2019ll end up processing the data in the same way for questions of the same type.<\/p>\n

\u00a0<\/span><\/h3>\n

Structure the Data<\/span><\/span><\/h3>\n

To make a professional-quality analysis often requires reshaping the data. The specific task we had with this survey was a one-time effort with the goal of producing a document, so for ease of use we used Power Query<\/a> and Tables in Excel to shape the data, and Power BI to create print visuals. If you face something larger, repeated, or more complex, it may be worth considering a more robust database solution<\/a> than Excel.<\/p>\n

Power Query\u2019s great advantage is that it makes a repeatable data transformation that can be refined and re-run, as opposed to a manual manipulation of the data that is difficult to repeat.<\/p>\n

These are some types of transformations to consider for surveys in particular:<\/p>\n