Personally, I find that survey data becomes burdensome somewhere around the 20th question. That applies to both paper and online surveys, though for different reasons. The first steps to making these large data sets manageable start all the way back to programming and setting up the survey. In this post, I’ll discuss four things I did that made sure I received my data in an organized fashion.
Open-ended questions in online surveys
The first fix is for online survey design.
Find every open-ended field where consumers are asked to enter a number (year of birth being the most frequent). Make sure you set restrictions on that field so your results are conforming.
By setting a requirement that the “year of birth” field be exactly four digits, you prevent people from entering their age instead, which could easily be confused for a two-digit date of birth (e.g., is “60” a 60 year old, or a 54 year old born in 1960?). This helps you to avoid the hassle of trying to clean a list of two digit birth dates or having to delete large amounts of data because you can’t discern whether it is an age or year.
Accurately categorizing paper surveys
I’ve mentioned in other blog posts that paper surveys that are being used to enter data should have a unique ID that associates them with the paper result. This allows you to compare your entered result to the actual document, letting you check for transcription errors.
You can help yourself further by making that unique identifier something relevant to the data. I’ve found by making my identifier either the date the survey was entered, or the date the survey was completed plus a letter (e.g. 0614a, 0614b, etc.) I can easily find the paper survey associated with any result.
Guiding readers through the survey
Include instructional text in every question where it is relevant.
While this does get tedious, it reinforces good habits in your ethnographer and makes sure that at no point is a multi-select mistaken for a single-select question, or vice-verse. This is particularly important for paper surveys.
Requiring an appropriate number of pages
Finally, make sure you condense a paper survey design to a singular page.
Having a second page doubles the amount of pages you receive and increases the possibility of lost data in the event the pages become separated. Should the pages arrive loose-leaf, all of your results could be invalidated if the pages become separated.
With one page surveys a stack of papers being knocked over is annoying. With multi-page surveys it could ruin an entire project.
Photo Source: https://www.flickr.com/photos/wilderdom/