Hi all,

here is the usual look at our most important readership metrics. This time examining, among other things, pageview changes in various countries, such as the effect of a brief block in China and of the sudden popularity of a Greek expression derived from Latin. The Android has been seeing a prolonged decrease in downloads, which fortunately were offset recently by a more positive development.

With this issue we are changing the scope of this report from a timespan of one week to two weeks (a fortnight) or four weeks, see further remarks below. Most of the metrics here are now being updated weekly on the new Product page.

Before we go to the usual data, a note (for those who haven’t seen it yet) that Wikistats has now been upgraded to the new pageview definition, see the announcement by Erik Zachte.

(All numbers below are averages for November 23-December 6, 2015 unless otherwise noted.)

Pageviews

Total:  527 million/day (-2.4% from the previous fortnight)

Context (April 2015-December 2015):

(see also the Vital Signs dashboard)

A notable weekly drop of -2.2% in the week until November 29, followed by another -0.2% in the week until December 6.

Desktop: 56.6% ​(week until Nov 22: ​57.2%)

Mobile web: 42.2% ​(week until Nov 22: 41.6%)

Apps: 1.2% ​(week until Nov 22: ​1.2%)

Context (April 2015-December 2015):

Global North ratio: 77.6% of total pageviews (week until Nov 22: 77.3%)

Context (April 2015-December 2015):

Out of curiosity about the -2.2% drop, I ran a query for the week until November 29 to find the countries with the largest changes from the previous week (as in some previous reports; restricted to those with >1 millions views/day in the week until Nov 29):

Greece +21.6% 2.3 m/day

South Africa -21.6% 1.2 m/day

Ireland   -20.6% 2.3 m/day

Colombia -15.7% 3.3 m/day

Venezuela -10.9% 2.3 m/day

Argentina -8.3% 4.9 m/day

Philippines -8.3% 3.5 m/day

New Zealand -7.3% 1.3 m/day

Vietnam -7.1% 2.1 m/day

Taiwan -6.9% 6.2 m/day

For the top three, I looked at how pageviews developed on a daily basis during the last three month including the week after this large change (until Dec 6):

In Greece, the +21.6% rise was the result of an isolated spike from November 23-25. This can be traced to a single page on the Greek Wiktionary which on most days before and after only saw a single-digit number of pageviews, but on these three days received more than 2.8 million: τάλε κουάλε. It’s about an expression that apparently comes from Latin via Italian (“tale quale”) and means something like “exactly the same” or “spitting image”. From the form of the spike, it was likely not the result of actual human interest, rather an undetected bot trying to learn exactly the same about exactly the same.

In Ireland, the -20.6% drop marked the end of a plateau whose start had actually shown up in the report for the week until November 1 already, where the country was the top changer with a 40.2% rise.

For South Africa, the -20.6% drop does not form part of a clear pattern.

Turning to the week until December 6, there were user reports starting around December 4 that Wikimedia projects were being blocked in China on desktop more widely than before (the Chinese Wikipedia has been blocked since May 2015). However, as can be seen in the chart below, the block appears to have been short-lived. It has also been mentioned in various media reports (English-language examples: China Digital Times, TIME).

New app installations

Android: 30.6k/day (-28.9% from the previous fortnight)

Daily installs per device, from Google Play

Context (last two months):

Instead of recovering to the level from before the App Store feature from Nov 5-12, installation numbers dropped further until early December. It now appears that this is connected to a sharp decrease in Google search referrals to the app’s store page. We’re looking further into it. The good news though is that the app is currently featured among the Best Apps of 2015 in various countries, with a clear effect on download numbers from December 3 on. (We’ll do a full assessment of the impact of this promotion in early January after it ends.)

iOS: 4.44k/day (-4.6% from the previous fortnight)

Download numbers from App Annie

Context (last three months):

There was a 5.4% drop between the week until November 22 and the week until November 29. But this is still higher than in the first week of November, say.

App user retention

Android: 14.8% (previous fortnight: 15.0%)

(Ratio of app installs opened again 7 days after installation, average of daily percentages for installs from the previous week. 1:100 sample)

Context (last three months):

(NB: In order to have the chart show any potential causal connections more clearly, I’ve changed the x-axis to the installation date instead of the date on which the survival is measured, and added annotations.)

iOS: N/A

(Ratio of app installs opened again 7 days after installation, average of daily ratios for installs from the previous week. From iTunes Connect, opt-in only = ca. 20-30% of all users)

Unfortunately I encountered data quality issues with this metric (again). We are in contact with Apple about this. I will leave the metric out of these reports until this is resolved or we have a suitable substitute (we could conceivably track this ourselves via EventLogging, like on Android).

Unique app users

Android: 1.158 million / day  (-3.8% from the previous week)

Context (last two months):

By now it looks like the App Store feature from Nov 5-12 indeed raised active users numbers a bit, although they have since been steadily decreasing, possibly for other reasons (see above).

iOS: 284k / day (+1.0% from the previous fortnight)

Context (last two months):

The week until November 29 saw an increase of 3.5% compared to the previous week. And in the chart, a little bump can be discerned from November 26 on, which could be related to Thanksgiving in the US (as we know, mobile use tends to be higher on weekends and presumably also holidays).

Schedule of this report

As mentioned last time, we have been rethinking the schedule of this report. New issues will now cover either two weeks or four weeks each (keeping it to a multiple of seven days, considering that most of these metrics show a strong weekly seasonality and comparisons are hence best taking that period into account). The main purpose remains the same as laid out earlier: to raise awareness about how these metrics are developing, call out the impact of any unusual events in the preceding week, and facilitate thinking about core metrics in general. Having iterated the selection and presentation of these metrics on a weekly basis since launching this report in September, they have now stabilized a bit. Also, it now appears that the newsworthy items uncovered by examining the development of these metrics can also be covered on a less frequent than weekly schedule. On the other hand, most of these metrics are now also updated weekly as part of the Reading team’s section on the new Wikimedia Product page on mediawiki.org. Feedback and discussion continue to be welcome.

----

For reference, the queries and source links used are listed below (access is needed for each). Unless otherwise noted, all content of this report is authored on behalf of the Wikimedia Foundation and released under the CC BY-SA 3.0 license. Most of the above charts are available on Commons, where I’m also uploading PDF versions of previous issues of this report.

hive (wmf)> SELECT SUM(view_count)/14000000 AS avg_daily_views_millions FROM wmf.projectview_hourly WHERE agent_type = 'user' AND CONCAT(year,"-",LPAD(month,2,"0"),"-",LPAD(day,2,"0")) BETWEEN "2015-11-23" AND "2015-12-06";

hive (wmf)> SELECT year, month, day, CONCAT(year,"-",LPAD(month,2,"0"),"-",LPAD(day,2,"0")) as date, sum(IF(access_method <> 'desktop', view_count, null)) AS mobileviews, SUM(view_count) AS allviews FROM wmf.projectview_hourly WHERE year=2015 AND agent_type = 'user' GROUP BY year, month, day ORDER BY year, month, day LIMIT 1000;

hive (wmf)> SELECT access_method, SUM(view_count)/14 FROM wmf.projectview_hourly WHERE agent_type = 'user' AND CONCAT(year,"-",LPAD(month,2,"0"),"-",LPAD(day,2,"0")) BETWEEN "2015-11-23" AND "2015-12-06" GROUP BY access_method;

hive (wmf)> SELECT SUM(IF (FIND_IN_SET(country_code, 'AD,AL,AT,AX,BA,BE,BG,CH,CY,CZ,DE,DK,EE,ES,FI,FO,FR,FX,GB,GG,GI,GL,GR,HR,HU,IE,IL,IM,IS,IT,JE,LI,LU,LV,MC,MD,ME,MK,MT,NL,NO,PL,PT,RO,RS,RU,SE,SI,SJ,SK,SM,TR,VA,AU,CA,HK,MO,NZ,JP,SG,KR,TW,US') > 0, view_count, 0))/SUM(view_count) FROM wmf.projectview_hourly WHERE agent_type = 'user' AND CONCAT(year,"-",LPAD(month,2,"0"),"-",LPAD(day,2,"0")) BETWEEN "2015-11-23" AND "2015-12-06";

hive (wmf)> SELECT year, month, day, CONCAT(year,"-",LPAD(month,2,"0"),"-",LPAD(day,2,"0")), SUM(view_count) AS all, SUM(IF (FIND_IN_SET(country_code, 'AD,AL,AT,AX,BA,BE,BG,CH,CY,CZ,DE,DK,EE,ES,FI,FO,FR,FX,GB,GG,GI,GL,GR,HR,HU,IE,IL,IM,IS,IT,JE,LI,LU,LV,MC,MD,ME,MK,MT,NL,NO,PL,PT,RO,RS,RU,SE,SI,SJ,SK,SM,TR,VA,AU,CA,HK,MO,NZ,JP,SG,KR,TW,US') > 0, view_count, 0)) AS Global_North_views FROM wmf.projectview_hourly WHERE year = 2015 AND agent_type='user' GROUP BY year, month, day ORDER BY year, month, day LIMIT 1000;

SELECT country_code, changeratio, ROUND(milliondailyviewsthisweek,1) AS milliondailyviewsthisweek FROM

   (SELECT country_code, ROUND(100*SUM(IF(day>22 AND day<30, view_count, null))/SUM(IF(day>15 AND day < 23, view_count, null))-100,1) AS changeratio, SUM(IF(day>22 AND day<30, view_count, null))/7000000 AS milliondailyviewsthisweek

   FROM wmf.projectview_hourly

   WHERE

     year = 2015

     AND month = 11

     AND agent_type = "user"

   GROUP BY country_code)

   AS countrylist

 WHERE milliondailyviewsthisweek > 1 GROUP BY country_code, changeratio, milliondailyviewsthisweek ORDER BY ABS(changeratio) DESC LIMIT 10;

https://console.developers.google.com/storage/browser/pubsite_prod_rev_02812522755211381933/stats/installs/ (“overview”)

https://www.appannie.com/dashboard/252257/item/324715238/downloads/?breakdown=country&date=2015-09-08~2015-12-06&chart_type=downloads&countries=ALL (select “Total”)

SELECT LEFT(timestamp, 8) AS date, SUM(IF(event_appInstallAgeDays = 0, 1, 0)) AS day0_active, SUM(IF(event_appInstallAgeDays = 7, 1, 0)) AS day7_active FROM log.MobileWikiAppDailyStats_12637385 WHERE userAgent LIKE '%-r-%' AND userAgent NOT LIKE '%Googlebot%' GROUP BY date ORDER BY DATE;

(with the retention rate calculated as day7_active seven days later divided by day0_active from the install date)

https://analytics.itunes.apple.com/#/retention?app=324715238

hive (wmf)> SELECT SUM(IF(platform = 'Android',unique_count,0))/7 AS avg_Android_DAU_last_week, SUM(IF(platform = 'iOS',unique_count,0))/7 AS avg_iOS_DAU_last_week FROM wmf.mobile_apps_uniques_daily WHERE CONCAT(year,LPAD(month,2,"0"),LPAD(day,2,"0")) BETWEEN 20151123 AND 20151206;

hive (wmf)> SELECT CONCAT(year,"-",LPAD(month,2,"0"),"-",LPAD(day,2,"0")) as date, unique_count AS Android_DAU FROM wmf.mobile_apps_uniques_daily WHERE platform = 'Android';