That is effectively the original plan. On Wed, Nov 6, 2019, 16:58 Georg Link <[email protected]> wrote:
> I like Ted’s proposed option 3. - Good thinking! > > Two identical surveys and we combine the responses for the analysis. > Survey 1: @apache emails, personalized invite links, no duplicate replies > allowed, response rate tracking > Survey 2: completely open for social media and other channels > > > > > On Nov 6, 2019, at 3:43 PM, Ted Dunning <[email protected]> wrote: > > > > On Wed, Nov 6, 2019 at 1:20 PM Georg Link <[email protected] <mailto: > [email protected]>> wrote: > > > >> > >>> On Nov 6, 2019, at 2:19 PM, Justin Mclean <[email protected]> > >> wrote: > >>> > >>> Hi Georg, > >>> > >>> Thanks for clearing up how that will work. > >>> > >>>> To send out invites through LimeSurvey, it requires creating a > >> “participant table”. > >>>> Every participant is assigned a token, to track whether the > participant > >> responded or not. > >>>> Again, the responses are anonymous and no link between the participant > >> entry and the response is stored. > >>> > >>> I assume we know if the participant responded but can’t match that to > an > >> individual response. That may not quite be as anonymous as some people > >> expect. > >> > >> To avoid tracking who responded I see two options, assuming we want to > >> send emails through LimeSurvey: > >> > >> Option 1: > >> We can send everyone the same URL with the same token from the dummy > >> participant. > >> > >> Option 2: > >> We can leave the survey open, without a participant table. Thus > >> eliminating the need for a token. > >> The invites are sent from a “dummy survey” that we don’t actually use > but > >> instead we include the URL to the real survey. > >> No tokens, no tracking, full anonymity, to information about response > rate. > >> > >> > > Option 3: > > Build a participant table and send with separate tokens per participant, > > but don't record the correspondence (that let's us avoid duplicate > > submissions for committers). > > At the same time, run a separate (but with identical questions) survey > for > > social media with no tokens to allow wide spreading. > > > > This allows us to compare the social media results against the > deduplicated > > committer results. That should make any spamming of the social media > branch > > fairly apparent. > >
