We have noticed that since we upgraded limesurvey from 1.92 to 2.06, the number of 'incomplete' responses we get seems to have increased significantly.

We have traced this to a change in behaviour of the system which is causing problems.

When a survey is accessed via a normal url with the surveyid, the Welcome screen loads and a record is not written to the database until the respondent clicks the button to begin the survey.

However, if a respondent accesses a survey via a link where there is an answer pre-populated on the URL via an SGQA parameter, limesurvey seems to create a database record immediately.

This is a problem because these 'incomplete' survey records can be generated in high volumes by bots following links like this that are posted online into forums or on websites etc.

The 'bots' are simply trawling and following links to index (or do whatever they do) the pages but in doing so they generate what appear to be 'incomplete' survey responses.

This can skew the apparent 'completion rate' of the survey and can consume unnecessary database resources.

I think the initial database write should not happen until at least the Welcome page (if there is one) has been displayed.

There is also a potential ethical problem here if limesurvey is capturing information (which may included ip address, refererurl etc.) before the respondent has been able to review any privacy terms that may be set out on the front page.