Collecting Metadata with Survey Software

Posted on : December 31, 2015 - by :

Metadata refers to all the information that can be collected during a survey other than the answers to the questions that appear.  The kinds of metadata you can collect depends on the capabilities of your survey software.

Most programs let you collect the date and time the survey was taken.  Somewhat more sophisticated programs let you also collect how long it took someone to complete the survey.  The most sophisticated software also lets you collect how long people took to answer each individual question.

If you are collecting your data online, you may be able to also collect the IP addresses of the people taking your surveys.  If you are collecting data using interviewers, you should be able to collect codes identifying which interviewer completed which survey.

Another metadata variable can be a completion code.  This would be a single number which tells you whether someone completed a survey or whether they were disqualified based on their survey answers or whether they were screened out, because they were over quota for a particular kind of person.

The main point of collecting all this data is to improve data quality.  For example, if you find that some surveys were completed much more quickly than you think reasonable, that would suggest the people taking them did not think about what answers they were giving.  If you find some surveys were taken at an odd time of day, that can mean they were taken by somebody in a time zone you are not targeting.  If you find the same IP address used many times to complete a questionnaire online, that may mean the same person is taking the survey over and over again.  In all these cases, you could eliminate the questionable data records.

And of course when using interviewers, you could use the interviewer ID’s to compare the results given by different interviewers and perhaps discover a problem with one or more of them.

In other cases you could add metadata to your analysis.  For example, if you are doing a political survey over several days or weeks, you could use the date the survey was taken to see if answers changed over time, perhaps due to some news item.

If your survey software lets you collect the time it took to answer individual questions, you could use that data to see if there are any questions that took an unusually long amount of time. If so, you could potentially rewrite those questions the next time you use them, making the survey process friendlier to your respondents.