search

mysociety / alaveteli-experiments

A collection of experiments and research to learn about what makes an Alaveteli grow
3 stars 0 forks source link

Write up of social media deep dive #86

Closed lizconlan closed 8 years ago

lizconlan commented 8 years ago

Closes #71

crowbot commented 8 years ago

So I'm convinced in reading this writeup and looking at the analytics, particularly 'Google Analytics - Conversions > Goals > Goal URLs > Make an FOI request (goal 1 completion)' that the /new URLs that have been shared are causing false positive goal conversions. I think therefore that it would be useful to add a section here looking at the figures once these requests have been excluded from this analysis. I've also created a ticket for recording the 'Make an FOI request' as an event rather than a URL, in order to avoid this problem in the future.

The time period used does not seem to be the one stated - '14 March - 13 June 2016' - it looks like it's '14 March - 13 April 2016'. Neither of these match the stated period on https://github.com/mysociety/alaveteli-experiments/pull/68, which is '16 March - 15 May 2016'. It would be good to match that period to make the relationship between the two bits of analysis clearer.

A couple of other bits of information that it would be useful to know:

crowbot commented 8 years ago

@lizconlan I guess I wasn't clear enough in my comments here - apologies for that - I was interested in seeing the data on who was tweeting and when the tweet was made (just after making request or once response has been received) for those same tweets in the first analysis, so that we could see if there was some relationship between conversion rate and who tweets/when they tweet. But given the questions around the recording of the conversion rate at the moment, I think it's going to be hard to draw any conclusions. Let's merge this writeup for now, and I'll move the ticket back to the contender list and make it dependent on getting a resolution merged for https://github.com/mysociety/alaveteli/issues/3360

lizconlan commented 8 years ago

when the tweet was made (just after making request or once response has been received) for those same tweets in the first analysis

I could have done better on that - there is an obstacle in the way that Twitter searches cut off (this article suggest it's 7 days but I thought it was a bit longer) preventing lookups of the t.co link code. A better approach would have been to used the Twitter's advanced search on their website to search for all tweets containing "asktheeu.org" on the day each tweet was made and go from there (bonus points for stats on retweets, I'd imagine). I got lost trying to work out how to get the data back