Research talk:Unique devices

Re: Cover all users

Do you really mean 100 %, or more like 95? Extreme examples: Internet Explorer 5 users, NoScript users, lynx users, ancient non-JavaScript mobile browsers. --Nemo 11:01, 1 December 2014 (UTC)

Yeah, users worth factoring in :(. We can do a pretty good estimate of the coverage (I still need to get around to it) but it won't be 100%. I'll write that in; thanks :). Okeyes (WMF) (talk) 12:55, 1 December 2014 (UTC)

1000 devices threshold

I see in [1] that there are only 424 rows, which means around 200 wikis, and there is no row under 1000 daily unique devices. Multiplied by 2*30, this means we're potentially excluding wikis with 60k unique visitors per month (worst case), is it really necessary? Nemo 18:56, 18 April 2016 (UTC)

wmfblog:2016/03/30/unique-devices-dataset/ claims «data for daily uniques was too sparse to be meaningful». Nemo 10:34, 23 December 2016 (UTC)

Thanks for the privacy conscious implementation

I just wanted to leave a public thanks for all the work and discussions that have gone into designing and implementing this counting method. I'm proud of the Analytics team and the rest of the Wikimedia developers who worked to find a stable and useful metric without resorting to the easy but highly privacy invasive methods that are common in the Internet industry. BDavis (WMF) (talk) 19:27, 19 April 2016 (UTC)

"Unique Devices" including Siri/Alexa/etc. usage?

Not sure if this is the right page to ask this question in, so please redirect me if I should ask elsewhere. I got here from this page on the awesome new Wikistats 2.0. Does the "unique devices" metric include usage of Wikidata or Wikipedia by "smart assistant" devices like Siri, Alexa, Google Home, etc. If it doesn't, do we have any usage numbers from those kind of devices, or are we even able to track that? --LiAnna (Wiki Ed) (talk) 18:36, 12 January 2018 (UTC)

I looked into this briefly - the general answer to the first question is most likely no. (According to Brandon from our Traffic team, Amazon appears to cache most requests for Alexa on their side, and the remaining requests are most likely counted as bots/spider traffic on our side, which is excluded from the Unique Devices metric anyway. Content for Siri appears to be retrieved using the Applebot user agent; these requests are marked with nocookies in our webrequest logs and thus likewise ignored as bot for the purposes of the Unique Devices metric. I assume Google Home works similarly.) Regarding the second question, I'm not aware of other usage numbers, but haven't looked. Regards, Tbayer (WMF) (talk) 01:57, 26 January 2018 (UTC)

PS: See now also wikitech:Analytics/Alexa (basically confirming the above mentioned hunch that Amazon caches most requests for Alexa on their side). Regards, Tbayer (WMF) (talk) 21:37, 8 February 2018 (UTC)

Alexa pulls the api once per page per edit so actual usage is very sporadic, given its user agent it is consider a bot and thus not counted. NRuiz (WMF) (talk) 21:43, 8 February 2018 (UTC)