Page MenuHomePhabricator

Create daily tracking reports for campaign: Banner campaign for Actionweek 2021
Closed, ResolvedPublic

Description

Tracking Part 1: banner campaign
Please provide daily reports during the banner campaign (08th of March to 21th of March).

Tracking Part 2: Landingpage links to Wikipedia
Please provide a tracking report for the Links from the Landingpage of the campaign until 8 weeks after the end of the campaign.

Timeline for the whole project

  • Start of the banner campaign: 08. March 2020
  • End of the banner campaign: 21. March 2020
  • Tracking test: first week of March
  • Last User signing up: 21.March 2020
  • Preliminary report for tracking part 1: sometime till end of March
  • Track editing behavior eight weeks after end of campaign: 02. Mai 2020
  • final report for tracking part 1 and 2: sometime till end of May

Banner Campaign information
There will be one campaign with 6 different banners (including mobile versions) which target readers in Wikipedia and logged-in user with up to 1.000 edits. Campaign goal are registrations for the newsletter of the community engagement team.

Find all details in this document.

[[central notice settings will be linked here]]
duration: 08th of March until the 21st of March 2021

Campaign Tags and Landingpages
Landing Page 1: https://www.wikimedia.de/wikipedia20/

Banner 1
?banner=WMDE_aw2021_var

Banner 2
?banner=WMDE_aw2021_ctrl

target groups

  • Banners will be targeted to non-logged in users
  • and registered users up to 1.000 edits.

There will be no A/B Testing.

Tracking is dependent on eventlogging . (https://phabricator.wikimedia.org/T250791 - ticket from 2020)

daily reporting during campaign
The following information should be included in the daily reporting

  • impressions per banner
  • clicks per banner/ page views per landing page
  • closing rate of banners

If possible with the distinction whether the person was logged in or not.

Event Timeline

There are a very large number of changes, so older changes are hidden. Show Older Changes

@GoranSMilovanovic I just saw that you claimed this ticket :)
Great - but i still need to complete it - so be patient. I hope I find the time early next week

@Merle_von_Wittich_WMDE No worries, I will follow the developments and take over once the campaign is ready for test and tracking.

@Merle_von_Wittich_WMDE @kai.nissen

TEST: event.wmdebannerinteractions; the userid field values are intentionally replaced by ---:

dt                      bannername              banneraction    bannerimpressions       userid
2021-03-03T10:56:24Z    WMDE_aw2021_ctrl        banner-seen     14                      ---
2021-03-03T10:54:37Z    WMDE_aw2021_ctrl        banner-clicked  15                      ---
2021-03-03T10:54:37Z    WMDE_aw2021_ctrl        banner-seen     15                      ---
2021-03-03T10:56:14Z    WMDE_aw2021_ctrl        banner-clicked  17                      ---
2021-03-03T10:54:47Z    WMDE_aw2021_ctrl        banner-closed   16                      ---
2021-03-03T10:56:14Z    WMDE_aw2021_ctrl        banner-seen     17                      ---
2021-03-03T10:56:24Z    WMDE_aw2021_ctrl        banner-clicked  14                      ---
2021-03-03T10:54:47Z    WMDE_aw2021_ctrl        banner-seen     16                      ---
2021-03-03T14:12:45Z    WMDE_aw2021_ctrl        banner-seen     18                      ---
2021-03-03T14:12:45Z    WMDE_aw2021_ctrl        banner-clicked  18                      ---

@Merle_von_Wittich_WMDE @kai.nissen

ATTENTION, PLEASE: TEST USER NOT FOUND FOLLOWING REGISTRATION. See T274563#6878638

SELECT year, month, day, hour,  event.campaign, event.userId, event.userName, webhost, wiki
      FROM event.serversideaccountcreation
      WHERE
        year = 2021
        AND month = 3
        AND event.campaign LIKE '%WMDE_aw2021%';

Outcome: an empty result set.

@GoranSMilovanovic

ATTENTION, PLEASE: TEST USER NOT FOUND FOLLOWING REGISTRATION. See T274563#6878638

I created a second one: Benutzer:Testuser actionweek2
Hopefully this time it works :) - could have been my cookie settings in FF

@Merle_von_Wittich_WMDE Thank you. We will have to wait until the event logging systems pick it up. I will be reporting back here as soon as I have something.

@GoranSMilovanovic I changed the banner keywords (bannername in the example you posted above) to reflect the actual campaign setup. There are two different campaigns, each running for one week and each containing banners for desktop, ipad and Android devices.

First week:

  • WMDE_wp20_aw_b1_desktop
  • WMDE_wp20_aw_b1_ipad
  • WMDE_wp20_aw_b1_mobile

Second week:

  • WMDE_wp20_aw_b2_desktop
  • WMDE_wp20_aw_b2_ipad
  • WMDE_wp20_aw_b2_mobile

@kai.nissen

Tested:

SELECT dt, event.bannerName, event.bannerAction, event.bannerImpressions, event.userID from event.wmdebannerinteractions WHERE event.bannerName LIKE '%WMDE_aw2021_%' AND year=2021 AND month=3 AND day=4;

Result set:

dt                      bannername              banneraction           bannerimpressions                    userid
2021-03-04T12:53:26Z    WMDE_aw2021_ctrl        banner-clicked         19                                   ---
2021-03-04T12:53:26Z    WMDE_aw2021_ctrl        banner-seen            19                                   ---

Q: @Merle_von_Wittich_WMDE has explained to me that the campaign tag WMDE_aw21 alone will be used from the landing page and onward; please let me know if there were any changes in that respect. Thank you.

@Merle_von_Wittich_WMDE As of T274563#6882513, the test user Testuser actionweek2 was not found:

SELECT year, month, day, hour,  event.campaign, event.userId, event.userName, webhost, wiki
      FROM event.serversideaccountcreation
      WHERE
        YEAR = 2021
        AND month = 3
        AND day = 4
        AND event.campaign LIKE '%WMDE_aw2021%' 
        AND wiki = 'dewiki';

returns an empty result set. In fact, there are no user registrations from the WMDE_aw2021 campaign in March 2021 at all.

Hey, do you have time for a short call today or tomorrow regarding your
questions/the tracking?

Am Samstag, 6. März 2021 schrieb GoranSMilovanovic <
no-reply@phabricator.wikimedia.org>:

GoranSMilovanovic added a comment. View Task
https://phabricator.wikimedia.org/T274563

@Merle_von_Wittich_WMDE
https://phabricator.wikimedia.org/p/Merle_von_Wittich_WMDE/ As of
T274563#6882513 https://phabricator.wikimedia.org/T274563#6882513, the
test user Testuser actionweek2 *was not found*:

SELECT year, month, day, hour, event.campaign, event.userId, event.userName, webhost, wiki

FROM event.serversideaccountcreation
WHERE
  YEAR = 2021
  AND month = 3
  AND day = 4
  AND event.campaign LIKE '%WMDE_aw2021%'
  AND wiki = 'dewiki';

returns an empty result set. In fact, there are no user registrations from
the WMDE_aw2021 campaign in March 2021 at all.

*TASK DETAIL*
https://phabricator.wikimedia.org/T274563

*EMAIL PREFERENCES*
https://phabricator.wikimedia.org/settings/panel/emailpreferences/

*To: *GoranSMilovanovic
*Cc: *kai.nissen, WMDE-leszek, GoranSMilovanovic, Aklapper,
Merle_von_Wittich_WMDE, Christine_Domgoergen_WMDE, Verena, Cirdan

@GoranSMilovanovic

Hello Goran,

I'm afraid that things are going wrong here because I don't know enough about tracking. Unfortunately I can't reach Christine (as my expert) because of the weekend.
17:00 CET ist perdect - I send you a hangout invite

But otherwise, here online for now

Tracking tester users:
I added the campaign tag after the normal URL on the landingpage. e.g.
//https://de.wikipedia.org/wiki/Wikipedia:Redaktionen?campaign=WMDE_aw21//
And I used one of those links to register a new account.

When I looked at your parameters you searched for "'%WMDE_aw2021%'" the used campaign tag on the landingpage is just WMDE_aw21.
I realise now, that I have not tested a user registration via banner --> landingpage --> onwiki.


Q: @Merle_von_Wittich_WMDE has explained to me that the campaign tag WMDE_aw21 alone will be used from the landing page and onward; please let me know if there were any changes in that respect. Thank you.

I think this is where we come to the trouble with the issue. Where I just don't have understanding of the tracking.
I'll break it down here, as I understood it.

  • The landing page can be reached in various ways: mainly via banners and social media advertising.
  • We have the campaign tag "WMDE_aw21" on the landing page, which is attached to all links that link to de.wikipedia pages. This tag is given to everyone who clicks on a link on the landing page that leads to Wikipedia.
  • the banners have their own tags (two banners, three different device types)

User Journey Steps and tracking see document

  • The first step of the user journey is covered by the banner tags
  • Step 2 and 3 are not a part of tracking
  • Step 4 of the user journey (see again the tracking document) is covered with the tag "WMDE_aw21"
  • Step 5: ??? I thought the tracking was also covered via the tag "WMDE_aw21" - but I realise now that especially considering users who already have an account, that this was probably a mistake on my part. – I guess this is our main topic for the call^^

@Merle_von_Wittich_WMDE

When I looked at your parameters you searched for "'%WMDE_aw2021%'" the used campaign tag on the landingpage is just WMDE_aw21.

% is the so called wildcard character in SQL queries, nothing to worry about, it just says that WMDE_aw2021 needs to be found in a string no matter how many other characters precede or follow it.

Also, I think that your understanding of the tracking concept for this campaign is correct.

Step 5: ??? I thought the tracking was also covered via the tag "WMDE_aw21" - but I realise now that especially considering users who already have an account, that this was probably a mistake on my part. – I guess this is our main topic for the call.

I thought the same thing, but since I do not implement any campaign tracking (I do not even know it works - @kai.nissen once provided a glimpse of it to me in Berlin) but just follow what happens in our databases, I do not think that I am the right person to ask.

See you at 17:00 CET, thanks for booking us a Hangouts session.

@Merle_von_Wittich_WMDE Thank you for catching this, user registration tagging works, of course - with the correct campaign tag!

year    month   day     hour    campaign        userid  username        webhost wiki
2021    3       3       14      WMDE_aw21       --- Testuser actionweek2021 de.wikipedia.org        dewiki
2021    3       4       12      WMDE_aw21       --- Testuser actionweek2    de.wikipedia.org        dewiki

@GoranSMilovanovic Hi Goran, would you share the spreadsheet with the daily reporting here as soon as you have it? Thank you!

@Christine_Domgoergen_WMDE

Hey Christine, of course, I will. I have just returned from my reimmunization from Covid-19 so I need to rest for a while. I will post a daily report in the evening. Thank you for your understanding.

@GoranSMilovanovic Yes of course, self-care first, especially in these times!! No worries, I am just always a bit curious about the numbers but it can easily wait until tomorrow or even the day after tomorrow. Take care!

@Christine_Domgoergen_WMDE @Merle_von_Wittich_WMDE @kai.nissen

Daily Report, 2021/03/08.

The banner interactions data:

Pageviews data:

  • no pageviews found; please check if a campaign user correctly picks-up the WMDE_aw21 tag from the landing page.

User registrations data:

  • nothing beyond two the already confirmed test users (see: T274563#6889229).

@GoranSMilovanovic @kai.nissen
Hi Goran, perfect, thank you! I had a look at the interactions table and am a bit confused: according to the table there were only 7709 impressions but 1014981 closing clicks and 519389 banner clicks. Could the impression number be wrong here?

There should be some pageviews already from my clicks only, @Merle_von_Wittich_WMDE maybe you can check the tracking tags again? Does Goran have the correct one?

@Christine_Domgoergen_WMDE

@GoranSMilovanovic has the correct tag. When he searches for WMDE_aw21 he should automatically find the unique ones, which have only added numbers with an underscore onto the WMDE_aw21.
But the tracking doc has also the unique tracking links included.

@GoranSMilovanovic What else could we try, that you can find the impressions?
There are probably not very much impressions, as there are only a few links and those are rather hidden. But for the two user registrations I also used the links with trackingtags from the landingpage and Christine also clicked on these links also rather heavily :D

@Christine_Domgoergen_WMDE @Merle_von_Wittich_WMDE @kai.nissen

T274563#6899474

according to the table there were only 7709 impressions but 1014981 closing clicks and 519389 banner clicks. Could the impression number be wrong here?

I didn't even check, we're using the same code to pick up the data from event.wmdebannerinteractions as we have used in the previous campaigns. I will now perform another search through that schema and monitor if there is something obviously wrong in the code, and there should not be anything. One thing that I have noticed is that the raw .tsv upon running a HiveQL query to fetch from event.wmdebannerinteractions resulted in an unusually large result set (857Mb).

T274563#6899529

When he searches for WMDE_aw21 he should automatically find the unique ones, which have only added numbers with an underscore onto the WMDE_aw21. But the tracking doc has also the unique tracking links included.

Then the analytics code should definitely find anything encompassing WMDE_aw21.

What else could we try, that you can find the impressions?

Definitely check if the campaign tag is passed on to our event logging databases. I can't help there. One more thing: do you remember exactly what pages where you visiting with any of the campaign tags attached?

Ok, I think I know what is the problem with banner interactions.

@kai.nissen Could you please confirm that a user who is not logged in will have a event.userID field value of zero? Thank you.

@Merle_von_Wittich_WMDE @Christine_Domgoergen_WMDE Do the following data on banner interactions look more plausible?

@GoranSMilovanovic Oh perfect, thank you! Banner impressions look a lot more plausible right now

@Christine_Domgoergen_WMDE

From T274563#6899924

I didn't even check, we're using the same code to pick up the data from event.wmdebannerinteractions as we have used in the previous campaigns. I will now perform another search through that schema and monitor if there is something obviously wrong in the code, and there should not be anything. One thing that I have noticed is that the raw .tsv upon running a HiveQL query to fetch from event.wmdebannerinteractions resulted in an unusually large result set (857Mb).

Well, I should have checked, because of the logged in vs. non logged in users distinction in this campaign. Ok, at least we have the banner interactions datasets in place now.

Let's see about those pageviews, please: could you remember what Wikipedia pages where you visiting from the landing page while carrying a campaign tag, or could you perform one such test today and let me know what URL exactly did you visit? Thank you.

@GoranSMilovanovic Okay, thanks! Can we distinguish in the number of impressions and clicks if they come from logged-in or non-logged-in users?

I just produced three page views here: https://de.wikipedia.org/wiki/Wikipedia:Wikimedia_Deutschland/LerneWikipedia?campaign=%20WMDE_aw21_7

Can you find them?

@Christine_Domgoergen_WMDE

Can we distinguish in the number of impressions and clicks if they come from logged-in or non-logged-in users?

Yes, we can. Would you like to see the distinction in the daily reports, or in the summary report? Because if it needs to be done for daily reporting I will also need to change the existing analytics code. It is easier done once we have all the data for the final report.

I just produced three page views here: https://de.wikipedia.org/wiki/Wikipedia:Wikimedia_Deutschland/LerneWikipedia?campaign=%20WMDE_aw21_7
Can you find them?

Testing now (EDIT. I mean: later during the day, we need to wait for pageviews extraction process from webrequests to finish first).

@Christine_Domgoergen_WMDE

Pageviews for today (more precisely: for what is found in wmf.webrquest until now):

"Tag","Page","Pageviews","date","campaign"
"?campaign=%20WMDE_aw21_7","de.wikipedia.org/wiki/Wikipedia:Wikimedia_Deutschland/LerneWikipedia",3,"2021-03-10","Actionweek_2021"
"?campaign=WMDE_aw21_1","de.wikipedia.org/wiki/Wikipedia:Edit-a-thon/Wirtschaftsuniversit%C3%A4t_Wien_zum_Internationalen_Frauentag_2021",1,"2021-03-10","Actionweek_2021"
"?campaign=WMDE_aw21_3","de.wikipedia.org/wiki/Wikipedia:WikiProjekt_Frauen/Frauen_in_Rot",1,"2021-03-10","Actionweek_2021"

So I was able to find (presumably) your: https://de.wikipedia.org/wiki/Wikipedia:Wikimedia_Deutschland/LerneWikipedia?campaign=%20WMDE_aw21_7

The problem is that, for some reason, in this campaign I have to use a somewhat different approach to pass the uri_path field to HiveQL.
Anyways: we have the pageviews data.

I will now re-run data collection for 2021/03/08 and 2021/03/09 and continue with a usual daily schedule as of tomorrow.

@Christine_Domgoergen_WMDE @Merle_von_Wittich_WMDE @kai.nissen

The campaign daily reporting sheet has been shared with you via Google Docs.

@Christine_Domgoergen_WMDE @Merle_von_Wittich_WMDE

Track editing behavior eight weeks after end of campaign: 02. Mai 2020

Please advise:

  • track editing behavior until 2021/05/02, or
  • track editing behavior eight weeks after the end of campaign?

@GoranSMilovanovic

Yes, we can. Would you like to see the distinction in the daily reports, or in the summary report? Because if it needs to be done for daily reporting I will also need to change the existing analytics code. It is easier done once we have all the data for the final report.

Oh good. No after the campaign for the interim report is fine!

Anyways: we have the pageviews data.

I will now re-run data collection for 2021/03/08 and 2021/03/09 and continue with a usual daily schedule as of tomorrow.

Great, thank you!

The campaign daily reporting sheet has been shared with you via Google Docs.

Perfect, that is very helpful, thank you.

Please advise:

track editing behavior until 2021/05/02, or
track editing behavior eight weeks after the end of campaign?

Thanks for noticing this. We would, like in previous campaigns, track the editing behavior up to six weeks after the campaign, so until 2nd of May.

@GoranSMilovanovic @gabriel-wmde @AbbanWMDE
Hello,
I just checked the tracking spreadsheet and I am confused by the banner names. Kai mentioned above

@GoranSMilovanovic I changed the banner keywords (bannername in the example you posted above) to reflect the actual campaign setup. There are two different campaigns, each running for one week and each containing banners for desktop, ipad and Android devices.

First week:

  • WMDE_wp20_aw_b1_desktop
  • WMDE_wp20_aw_b1_ipad
  • WMDE_wp20_aw_b1_mobile

Second week:

  • WMDE_wp20_aw_b2_desktop
  • WMDE_wp20_aw_b2_ipad
  • WMDE_wp20_aw_b2_mobile

Yet in the tracking sheet there are the following banner names for the first day (8.3.2021):

WMDE_wp20_aw_b1_desktop
WMDE_wp20_aw_b1_desktop_ipad
WMDE_wp20_aw_b1_desktop_mobile
WMDE_wp20_aw_b1_ipad
WMDE_wp20_aw_b1_ipad_ipad
WMDE_wp20_aw_b1_ipad_mobile
WMDE_wp20_aw_b1_mobile
WMDE_wp20_aw_b1_mobile_ipad
WMDE_wp20_aw_b1_mobile_mobile
WMDE_wp20_aw_b2_desktop
WMDE_wp20_aw_b2_desktop_ipad
WMDE_wp20_aw_b2_desktop_mobile
WMDE_wp20_aw_b2_ipad
WMDE_wp20_aw_b2_ipad_ipad
WMDE_wp20_aw_b2_ipad_mobile
WMDE_wp20_aw_b2_mobile
WMDE_wp20_aw_b2_mobile_ipad
WMDE_wp20_aw_b2_mobile_mobile

This confuses me because of:

  • B2 shouldn't have had any impressions in the first week, no?
  • Every banner (desktop, mobile, ipad) seems to exist in three versions again (desktop, mobile, ipad) --> which is the correct marker for the device where it was displayed?

Then the banner names in the sheet change on the second day:

WMDE_wp20_aw_b1_desktop
WMDE_wp20_aw_b1_desktop
WMDE_wp20_aw_b1_desktop
WMDE_wp20_aw_b1_desktop
WMDE_wp20_aw_b1_desktop
WMDE_wp20_aw_b1_desktop
WMDE_wp20_aw_b1_desktop
WMDE_wp20_aw_b2_desktop
WMDE_wp20_aw_b2_desktop_ipad
WMDE_wp20_aw_b2_desktop_mobile
WMDE_wp20_aw_b2_ipad
WMDE_wp20_aw_b2_ipad_ipad
WMDE_wp20_aw_b2_ipad_mobile
WMDE_wp20_aw_b2_mobile
WMDE_wp20_aw_b2_mobile_ipad
WMDE_wp20_aw_b2_mobile_mobile

  • why do they change? because for the banner names missing there were no impressions in the database?
  • why is there less impressions on B1 although this should be the main banner?
  • why is there only b1_desktop and no mobile or ipad?

So it goes on like this and there are different combinations of banner names in the spreadsheet every day. Could you have a look and find out why this happens? Is it something in the tracking code or if the schema isn't working correctly? Thank you!

@Christine_Domgoergen_WMDE

All banners in this campaign contain _wp20_aw_ somewhere.

The analytics code which produces the datasets presented in the Campaign Daily Reporting Spreadsheet just looks for _wp20_aw_ anywhere in the campaign tag attached from interacting with any particular campaign banner.

In effect, what you see in the daily reporting spreadsheet are simply all banners that contained _wp20_aw_ on a particular day.

What banners exactly where used and when is now beyond my knowledge since this complication that you have described has emerged.

It would be also beneficial for me to learn about the exact organization of banners in the first and the second part of the campaign so that I can organize our analyses appropriately.

@GoranSMilovanovic Okay, thank you!

@gabriel-wmde @AbbanWMDE Can you help with that? This is quite important for our analytics... thank you!

@GoranSMilovanovic can you continue the daily reporting anyways? I guess it will take a bit to figure out the confusion and we might need @kai.nissen for that...

Phase 1: completed:

Tracking Part 1: banner campaign
Please provide daily reports during the banner campaign (08th of March to 21th of March).

Phase 2: begins.

Tracking Part 2: Landingpage links to Wikipedia
Please provide a tracking report for the Links from the Landingpage of the campaign until 8 weeks after the end of the campaign.

Q. Just to confirm: we will not be tracking any user edits here?

Hi @GoranSMilovanovic,

For any campaign-caused registration the edits should be tracked. This includes registrations through the Landingpage.

@Verena

But I don't have the data on user registrations specifically from the landing page... if that is possible to have at all: if a user went to registration straight from the landing page, did they carry the campaign tag with them, or not? I am not sure. Please advise. Thank you.

To clarify: all registered users that are found in our databases had a campaign tag attached before registration. So, if that happens from the landing page, then they are in our databases, if not... well, then they are not.

No worries. For the landingpage the campaign tag is ?campaign=WMDE_aw21 Any entries in the database related to that tag should be included in the report.

@Verena Nice! We can then track all registered users' edits.

The preliminary report is under preparation. User edits tracking begins today.

Thank you.

@Verena Preliminary findings: thus far, four out of 17 campaign registered users made edits. We will continue to track user edits until 2nd of May as planned (sorry for introducing confusion in T274563#6934320 and thanks for your clarification!)

@GoranSMilovanovic Perfect, thank you! When will you be able to finish the preliminary report?

@Christine_Domgoergen_WMDE

The preliminary report will ve ready and shared today.

@GoranSMilovanovic

Great, thank you! I have some questions:

  • Could you add the analysis category of logged in users vs non logged in users for for banner impressions, banner clicks and closing clicks? Thank you.

Chart 0.1.7 Daily Banner Click and Close rates, by b1/b2, aggregated across the campaign banners.

  • Why do we have 376149 banner impressions of banner 1 on March 8th when the banner should be shown only in week two? Was the banner live accidentally or might there be a mistake in the database? I guess we need @kai.nissen here, can you have look Kai?
  • Why do we have a small number of banner impressions of banner 2 in the second week and quite a high number of banner impressions of banner 2 on March 15th?

Chart 0.1.6 Daily Banner Click and Close rates, by devices, aggregated across the campaign banners. and Chart 0.1.6B Mean, median, and total Impressions, clicks and closing clicks per banner , per device. NOTE. Log(y) scales were used to help readability; data labels are exact values.

  • I am still confused with the banner names/devices here, as mentioned in the previous comment. We need the banner impressions, clicks and closing clicks per device. From the current banner names I cannot calculate these numbers (what is desktop_mobile, desktop or mobile?). @kai.nissen Could you explain how to interpret the banner names? And when we have the explanation could you Goran incorporate this in the report?

@Christine_Domgoergen_WMDE

Could you add the analysis category of logged in users vs non logged in users for for banner impressions, banner clicks and closing clicks? Thank you.

Done: the report is too large for Phabricator and will be shared via e-mail.

Sorry for the late reply! I'll try to de-confuse: The banner code still contains suffixing based on the viewport width. I actually didn't expect it to be there, but it made its way into the production banners.

The banner names as hardcoded in Central Notice are (same for the pre-campaign week banners (b2)):

  • WMDE_wp20_aw_b1_desktop
  • WMDE_wp20_aw_b1_ipad
  • WMDE_wp20_aw_b1_mobile

Depending on the viewport width the banner code appends _ipad or _mobile to the banner names for use in Event Logging and Matomo campaign parameters. It does not append _desktop when the banner is displayed on larger screens.

That has different implications on how to read the numbers:

suffixexplanation
desktopdesktop device
desktop_ipaddesktop device with a smaller browser window (or screen)
desktop_mobiledesktop device with a smaller browser window (or screen)
ipadipad device with a larger screen
ipad_ipadipad device with the expected screen size
ipad_mobileipad device with a smaller screen size
mobileiPhone or Android device with a larger screen (probably mostly Android tablets)
mobile_ipadiPhone or Android device with a larger screen (probably mostly Android tablets)
mobile_mobileiPhone or Android smartphone

Although this feels like an error in tracking, it gives us a deeper insight by showing which banner version (in terms of responsive design) the users actually saw. For instance, quite a lot of people clicked the iPad version of the banner on their desktop device. And we get a hint on how many people used an Android tablet when clicking/seeing the banner.

If it is the device type that we are most interested in, the additional suffix should be removed. If we are more interested in which banner the users actually saw, we should remove the first suffix, but add desktop if there is only one suffix.

@kai.nissen Great thank you!

@GoranSMilovanovic For the report this means we could make the following adjustments:

Chart 0.1.6B Mean, median, and total Impressions, clicks and closing clicks per banner , per device.

  • Could you calculate the total click and closing rates for each banner in this chart (as in Table 0.1.5 De facto Click and Close rates in percents)?

And could you summarize banner impressions, clicks and closing clicks per device by grouping the banners together like this:

suffixexplanationanalysis category
desktopdesktop devicedesktop
desktop_ipaddesktop device with a smaller browser window (or screen)desktop
desktop_mobiledesktop device with a smaller browser window (or screen)desktop
ipadipad device with a larger screenipad
ipad_ipadipad device with the expected screen sizeipad
ipad_mobileipad device with a smaller screen sizeipad
mobileiPhone or Android device with a larger screen (probably mostly Android tablets)ipad
mobile_ipadiPhone or Android device with a larger screen (probably mostly Android tablets)ipad
mobile_mobileiPhone or Android smartphonemobile

Thank you!

@GoranSMilovanovic Hi Goran, I will be on vacation for the next weeks but @Merle_von_Wittich_WMDE will take over from here and looks forward to the updated report :-)

@Christine_Domgoergen_WMDE @Merle_von_Wittich_WMDE @WMDE-leszek

I am currently prioritizing some Wikidata tasks because the deadline for this report is nowhere near, but no worries I am on it and it will be done in the following days.

@Christine_Domgoergen_WMDE Enjoy your time!

@Christine_Domgoergen_WMDE @Merle_von_Wittich_WMDE @WMDE-leszek

I am currently prioritizing some Wikidata tasks because the deadline for this report is nowhere near, but no worries I am on it and it will be done in the following days.

@Christine_Domgoergen_WMDE Enjoy your time!

@GoranSMilovanovic
Hey, I just wanted to ask if there is anything new here yet: we want to post the evaluation on the German Wikipedia project page soon.
If you are still busy with Wikidata, then we will post the analysis in a shortly and add a note that the report will be updated at a later point in time.

@Merle_von_Wittich_WMDE @WMDE-leszek @kai.nissen

The report is shared via e-mail.

In reference to T274563#7027683 please take a look at tables 0.1.6C and 0.1.6D.

The user edits data were also updated but it does not seem that anything interesting is happening there.

@Merle_von_Wittich_WMDE @WMDE-leszek @kai.nissen

The report is now update to check for user edits until 02. May 2021 as demanded in the ticket description.

The report will be shared via e-mail.

Please let me know if there is anything else needed here.

@Merle_von_Wittich_WMDE @kai.nissen @WMDE-leszek

Please, do we need any additional work here? Thanks!

Hey @GoranSMilovanovic I do not think so :), but Christine is back in the office next week and I would like to double check with her, if that is fine with you?

but Christine is back in the office next week and I would like to double check with her, if that is fine with you?

Absolutely fine.

@GoranSMilovanovic Thanks for the updated reports and your work! I think we have everything we need and I will close the ticket now :-)