Do you look at non-anonymized user data in your analytics and tracking?
No—we don’t look at non-anonymised user data in our analytics. We use Google Analytics events, so we can see e.g. a graph of how many forecasts are made each day, and this tracks the ID of each user so we can see e.g. how many users made forecasts each day (to disambiguate a small number of power-users from lots of light users). IDs are random strings of text that might look like cwudksndspdkwj. I think you’d call technically this “pseudo-anonymised” because user IDs are stored, not sure!
Who specifically gets access to user submitted predictions (can’t quite tell how large your team is, for instance)
Your predictions are private to you unless you share them. I and the other two devs who have helped out with parts of the project have access to the production database, but we commit to not looking at users’ questions unless you specifically share them with us (e.g. to help us debug something). I am interested in encrypting the questions in the database so that we’re unable to theoretically access them, but haven’t got round to implementing this yet (I want to focus on some bigger user-visible improvements first!)
Hope this makes sense! Thanks for your kind words and for checking about this, let me know if you think we could improve on any of this!
By the way, very tiny bug report: The datestamps are rendering a bit weird? I see the correct date stamp for today under the date select, but the description text in italics is rendering as ‘Yesterday’, and the ‘data-tip’ value in the HTML is wrong.
Obviously not a big deal, just passing it on :) I’m currently in PST time, where it is 9:39am on 2023.07.25, if it matters. (Let me know if you’d prefer to receive bug reports somewhere else?)
Thank you!
No—we don’t look at non-anonymised user data in our analytics. We use Google Analytics events, so we can see e.g. a graph of how many forecasts are made each day, and this tracks the ID of each user so we can see e.g. how many users made forecasts each day (to disambiguate a small number of power-users from lots of light users). IDs are random strings of text that might look like
cwudksndspdkwj. I think you’d call technically this “pseudo-anonymised” because user IDs are stored, not sure!Your predictions are private to you unless you share them. I and the other two devs who have helped out with parts of the project have access to the production database, but we commit to not looking at users’ questions unless you specifically share them with us (e.g. to help us debug something). I am interested in encrypting the questions in the database so that we’re unable to theoretically access them, but haven’t got round to implementing this yet (I want to focus on some bigger user-visible improvements first!)
Hope this makes sense! Thanks for your kind words and for checking about this, let me know if you think we could improve on any of this!
Thanks for the fast response, all of this sounds very reasonable! :)
By the way, very tiny bug report: The datestamps are rendering a bit weird? I see the correct date stamp for today under the date select, but the description text in italics is rendering as ‘Yesterday’, and the ‘data-tip’ value in the HTML is wrong.
Obviously not a big deal, just passing it on :) I’m currently in PST time, where it is 9:39am on 2023.07.25, if it matters. (Let me know if you’d prefer to receive bug reports somewhere else?)
Ah thank you! I’ve just pushed what should be a fix for this (hard to fully test as I’m in the UK).
Thanks so much! :) FYI that the top level helper text seems fixed:
But the prediction-level helper text is still not locale aware:
(Again, not a big deal at all :) )