Michael Sinner

Digital Analytics Consultant

Track Ezoic Ad-Exposure via GTM for more Transparancy

Ezoic Ads has quite a success story and became quickly the Monetization-Blogosphere’s darling. With AI to boost your ad moneys from websites you build mainly for monetization reasons. There are all the great reviews, showcases and stories you’ll find on the web. Like on nichepursuits.com where it all started for me, when I first heard of Ezoic. Quickly I realized there is one major concern I have with AI in general that just bubbled up when I implemented Ezoic the first time in a WordPress blog:

  • AI is intransparent in its decisions and outcome (in this case of automized ad display, as every client will have a different amount of ads displayed, and therefor we have no clue what exactly is the result)

Unfortunately the human nature (at least this applies to myself) is curios. I need to know what why and how something happened. Can you just rely on the Ezoic systems that there was „just the right“ amount of ads displayed to your users? How many ads were displayed per user on a single page? Which ad-spaces were empty and which not? Is my content overcrowded with ads?

Lets find out. By implementing some code and track the s**t out of all the ads, that are displayed to my websites visitors ­čÖé
(no, not on this page though; no moneys for me here)

How many ads are optimal?

Lets first dig into the actual motivation on this post. It’s not all just about the knowledge how many ads my users see. Sure, thats really interesting to know, but whats the business case?

To answer the question of optimal ads per user or per page, you will quickly realize its not 42. No. Not even a certain number. Its a rather a rate of number of ads per user per time. Often a blogger will draw a line of 6, 4 or some other Number which they feel comfortable with. However thats too much of personal bias and so on. Which I found myself in, when I first thought about adding display ads to one of my blogs. Users mustn’t see too many ads and must be able to access the content first. As I put a crazy amount of time and effort to create content.

Source: ezoic.com

There actually is a huge post on this whole optimal ad count topic on Ezoic’s blog. Yes, it is actually a very good read with lots of insights that is way more than a „lets put some thin content so we have something“-blogpost: How Many Ads Are Too Many?

And with this I’ll close this little side-topic without digging deeper. Lets focus on the main question at hand.

How can I track the amount of ads displayed with Ezoic in GTM?

With Google Tag Manager (GTM) there is a little framework I installed to first analyze the amount, size and meta of ads displayed on a page to a client. This first check creates a „report“ and pushes all the information back to my Data Layer. Second I populate all the information to Google Analytics (GA) via Enhanced Ecommerce Tracking and Internal Promotion Tracking. This way I take advantage of native reports in the GA web interface (well, if you actually want to use this…). But even better: this includes a batching of all the data into a single request. As we will see, the amount of data tracked can quickly explode here, and we do wan’t batches. Right?

Very First: Challenges and Caveats

Now before we get into the code. You must know this. We will built a code that is prone to fail, as we anchor all our information on CSS selectors and HTML DOM information that Ezoic dynamically renders into a page. There are certain templates Ezoics code uses to embed the ads. But those will change as they update and improve their templates. And then this GTM code will fail. I have no clue when this will happen and how much effort there will be to maintain this code in GTM. I’ll keep you updated though, as I see something break in my data and code.

There is also one more challenge we will have to face. The dynamic nature of ads that Ezoic beds into your pages. Not all ads are displayed exactly on page load, ready or any other state. Even after some time the ads are reloaded, updated and empty spaces are filled and filled spaces are emptied. This challenges our tracking, but we’ll get to this in a moment.

First: Analyze the amount of ads displayed

Create a HTML tag in your GTM. The following code will analyze a pages DOM and look for ezoic ads. All ad spaces that are found are populated into an array of objects. With this properties:

  • index: a simple counter per ad space
  • width: the HTML objects property „scrollWidth“ which tells us the actual size
  • height: same as width, you get it
  • placement: the HTML objects property from the dataset directly on exactly the object found
  • comments: this contents all the HTML comment information from the object found. Including the objects parents comments and all further comments inheritated. This is implemented because Ezoic facilitates HTML comments to push meta information on ads that is very interesting and useful.

Second: React and Read the Data

To facilitate the data we got from the „Transparancy Report“ HTML tag we need some basic „groundwork“ GTM variables.

Basic Variables (Data Layer Variables):

  • dataLayer.ezoic.adImpression (simply map this Data Layer Var to „ezoic.adimpression“ which is populated in the tag above)
  • dataLayer.ezoic.submittedAdsCounter (simply map this Data Layer Var to „ezoic.submittedAdCounter“)
  • static.ezoicAdsCssSelector: This is┬áthe „core“ item which is used to identify ad placements from Ezoic. Prone to fail, as this will change over time. As of now (2020-08-05) the content is:┬á[id*=ezoic-pub]

More complex Variables:

These three JS Variablers we will need in a second, just bear with me.

JS Variable Name: „js.ezoicEecPromotionImpression“

(this will create an object that is ready to use in GA’s EEC tracking)

JS Variable Name: js.ezoicVisibleAdsCounter

(this will use the same CSS selector as the main ‚Transparancy Report‘ HTML tag and simply┬ácount┬áthe visible ads)

JS Variable Name: js.ezoicAdUpdated

(this returns whether the count of visible ads is greater than the last known counter when ad-impressions were submitted to GA)

The Timing Challenge

As already mentioned there is a major challenge that we still need to tackle. Brace up, as we discuss some ideas on how to overcome this. The ever so dynamic nature of Ezoic ads. Which just can’t stay still and change in their size and content, after load and from second to second as the client browses a page.

Mainly because of the GTMs queueing system, we interact in loops of events that just so happen all the time. Page loaded. DOM ready. A click, a scroll and so on. But there is no single clear event on when an ad is loaded, displayed or updated. So we again have to build some adapter workaround countermeasure, which again comes with all the disclaimer. One more time I ask you to build error prone code.

There are two options I see to work around this issue.

  1. Facilitate mutation observer events: use the CSS selector we already have to facilitate mutation observers as an API to trigger the ‚Transparancy Report‘ tag.
  2. Use static triggers and events you already have and check if Ezoic ads changed between two events.

The first approach is much more detailed and will give you the most insights. However this comes with the cost of added complexity (adding hazard of shaky code), browsers compatibility issues and performance that will increase the load on your clients browser. Therefor I didn’t implement any such solution and will showcase just the latter approach.

Instead I use the following system to trigger my Ezoic ads information loop. So the main GTM Tag to analyze and report what ads are displayed has two triggers actually.

a)┬áSome pretty early event (such as Pageview, DOM Ready,…) where hopefully some ads are already visible.

b) Any event, later then the first triggering event, where the amount of visible Ads changed.

This is as tricky as its vague in description. The reason is, that the very first ads will probably load way later than the actual page is displayed to the user. Therefor we need the variable, which simply gives us an Integer counter of how many ads are visible at this exact event in our GTM queue (dataLayer). The first (and initial) transparancy report mustn’t start prior this counter is >0. Otherwise we will bloat our GA data with empty-ad-space information. Though maybe its also helpful, so it’s also possible to just track an initial status (without considering the counter). This is absolutely open for discussion and to be decided per individual case.

Then there is this second trigger, which is very vague described as „any event, where ad count changed“. This setup will only work, if you already know there are more events to come on every relevant pageview. Or you’ll have to implement a timer trigger or similar.

In my case there actually is a 10sec trigger already installed to track basic interaction rates with my blogs content. Just a short check if the user already dropped out or is still there, reading the content. Also there are more clicks and interactions tracked, so I generally assume there are „enough“ events to be queued and trigger my Transparancy Report tag. This is the actual trigger:

For some reason the GTM interface won’t keep this english and always translates the text to german; even though the rest is english

And thats all the workaround to give an educated guess what goes on with my page and all the ads. Now what to do with all this data?

Finalize with a pinch of EEC Tracking

The data streams in┬áthe ad information pipeline. As the pages load and with every event the system checks for changes and reports if something happened. If so we receive dataLayer updates. With an array full of ad display objects and properties. Now arrays with multiple objects are a multidimensional thing but GA in its basic setup only expects two dimensional observations („a [pageview|session|event] happend with the following properties and values“).┬áEEC to the rescue!

Fortunately around 2014 already GA took care of this. With enhanced ecommerce and the option to fill index based observations into single dimensions (like a list of products or internal promotions). Which is exactly what we need to track multiple ad observations on a „flat“ single pageview interaction.

Therefor I created a little translation adapter script, which will take the Ezoic ad information and parse it into the EEC internal promotion JSON representation. This is the GTM variables code (custom JS variable): „js.ezoicEecPromotionImpression“ (see above)

This is the final configuration of my GA event.

Now what will this look like in GA when all this GTM code is published onto production?


In this post I created the following steps to analyze the ad exposure to website clients with GTM.

  1. Create the CSS selector based HTML/JS tag, which check the page for ads. These observations are pushed into the dataLayer.
  2. Create meaningful triggers that work for your case, where I only described some options and caveats.
  3. React to the transparancy reports from the first step and send the data to GA, e.g. with GA enhanced ecommerce and internal promotions.

Google Analytics / Universal Analytics HTTP/URL Parameter Liste

Google Analytics ├╝bermittelt mittels Measurement Protocol die Daten, was auf HTTP basiert. Das bedeutet man kann mit beliebiger Software den Verkehr im Browser und auch in der App abfangen und analysieren. Hier eine kurze ├ťbersicht ├╝ber alle Parameter im HTTP Request zu google-analytics.com/collect (Measurement Protocol) und /__utm.gif (klassisches Analytics). (mehr …)

Google Tag Manager – Auto Event Tagging

Google Tag Manager Logo

Eine der wesentlichen Neuerungen an Google Analytics bzw. Google Tag Manager (GTM) die beim Google Analytics Summit 2013 vorgestellt wurden, war das Auto-Tagging. GTM erlaubt zwar Tags einfach in die Seite einzubauen ohne technische Ressourcen zu beanspruchen, aber Ereignisse bzw. Events m├╝ssen nach wie vor erst implementiert werden. Und das bedeutet wieder einen Change Request an der Website, der von der Agentur oder sonstigen Dienstleistern erst mal eingebunden werden muss. Das Auto Event Tagging erm├Âglicht uns nun mit viel weniger Aufwand in sehr kurzen Intervallen Events zu erfassen. Hier eine kurze Anleitung. (mehr …)

Google Analytics Summit 2013 Mountain View – AllesInAllem

google analytics summit babak pahlavan

Baba Pahlavan stellt die 14 Announcements vor

Alles in Allem war auch dieses Jahr der GA Summit bei Google eine riesen Party, bei der ein Feuerwerk gez├╝ndet wurde. Nicht so gro├č war es und stand etwas im Schatten der letzten Jahre, da es keine Ank├╝ndigung wie Universal Analytics gab. Dennoch unheimlich spannend und an manchen Stellen PipiInDenAugen-Momente. Content Channeling, endlich! Account Management Write API – super! (mehr …)

Mobile und Google – wieso vertr├Ągt sich das so gut?

Google Mobile Studien Logo

Google Mobile Studien sollen aufkl├Ąren, verraten aber nicht den wahren Hintergedanken des Konzerns.

76,5% Wachstum in Mobile Ads, 97% aller Smartphone-Besitzer nutzen das Internet, 92% benutzen Suchmaschinen mobil. Solche Fakten werden uns von Studien und Google um die Ohren geschmissen. Mobile ist das n├Ąchste gro├če Ding! Ok, wenn es zumindest nach den Konzernen geht. Vertrauen funktioniert durch Studien, denn diese sind augenscheinlich objektiv und werden von unabh├Ąngigen Instituten erhoben. Doch wieso steckt Google und Mobile so sehr unter einer Decke? (mehr …)

Mobile vs. Desktop Nutzung – Infografik

Vergangenes Jahr (2012) habe ich eine Masterarbeit zum Thema „Reichweitenerh├Âhung von Nachrichten- und Medienportalen durch Einsatz mobiler Anwendungen“ verfasst. Die wesentliche Studie bestand aus der Auswertung von mobile und non-mobile Zugriffen auf die Websites meiner Probanden Spiegel.de, Welt.de, Abendblatt.de und Morgenpost.de. In Kooperation mit den Websites habe ich interne Daten aggregiert und ausgewertet.

Das Ergebnis m├Âchte ich etwas verz├Âgert heute in Form einer Infografik und des (mehr …)

Facebook und Google schaffen universelle Cookies

Pl├Ątzchenzeit… ist vorbei. Nicht aber f├╝r Facebook und Google. Letzterer k├╝ndigte bereits Ende 2012 an eine universelle Google ID auf den Markt zu bringen. Facebook hat diese Ank├╝ndigung erst vor wenigen Tagen gebracht. Wir werden ├╝berall erkannt und ├╝berall werden unsere Daten an die zentralen Datensammelstellen ├╝bertragen. Quasi eine Institution f├╝r Daten, sie wissen mehr ├╝ber uns als unsere Kinder es je erfahren werden. (mehr …)

Printkrise: Inhalte und Medium synchronisieren

Bei Gruner+Jahr zu arbeiten ist eine tolle M├Âglichkeit f├╝r mich um hautnah zu erfahren, wie es in der Printbranche derzeit drunter und dr├╝ber geht. Selbst wenn man das meiste nicht intern erf├Ąhrt, sondern durch alle anderen Kan├Ąle mitgeteilt bekommt. Die Financial Times Deutschland und die Frankfurter Rundschau geht vor die Hunde, so dem Titel der aktuellen Zeit wohl nach zu urteilen. Wo ist denn da Sand im Getriebe?

titel der ZEIT

Die ZEIT diese Woche, wie kann Journalismus ├╝berleben?

 Medien im Wandel

Nat├╝rlich ist nicht nur alleine der Wandel der Medien einziger Ausl├Âser f├╝r die Probleme der genannten Printtitel. Wer rote Zahlen schreibt wird ├╝ber Kurz oder Lang immer den k├╝rzeren ziehen. Doch gerade in diesen Zeiten hat kein Verlag mehr die Mittel um einen Klotz am Bein zu subventionieren. Bei Gruner+Jahr ist n├Ąmlich Sparkurs auch bei Brigitte angesagt, 17 Mitarbeiter weniger wird man demn├Ąchst in der Redaktion haben. Die Einnahmen durch klassische Werbung sehen heute aus wie ein Flickenteppich, ├╝berall entstehen L├╝cken. Wieso k├Ânnen diese L├╝cken in den Verlagen nicht gestopft werden? Ist es nicht allgemein bekannt, dass die Werbeetats eine konstante Entwicklung aufzeigen?

Es ist nichts Neues

Schon lange bekannt ist doch, dass man sich auf neue Technologien erst einmal einstellen muss.┬áÔÇ×[…]journalism practices for new media have typically departed from editorial workersÔÇÖ perceptions and experiences with which news media journalists are familiar (the old). When radio was introduced, journalists initially read newspaper articles aloud in radio broadcasts. When television news broadcasts began, journalists made radio broadcasts, but added moving images.“ – Oscar Westlund 2011 „Sensemaking of the Mobile Media (R)evolution

Wenn man das auf heute ├╝bersetzt, dann sieht man sofort die Analogien. Erst hat man einfach die Printprodukte eins zu eins ins Internet ├╝bertragen, bevor man auf die Idee gekommen ist, dass das Internet nichts statisches ist. Das war vor bald 20 Jahren. Und wo stehen wir heute? Haben wir alle das Gef├╝hl, dass die Verlage das digitale Medium beherrschen? Wenn ich mir so die Kommentare zu den aktuellen Artikeln auf SPON, Zeit, etc. ansehe, dann sind wirklich alle gro├čen Titel noch sehr weit davon entfernt, das Internet so zu nutzen, was seine M├Âglichkeiten hergeben.

Aber selbst das wird die Zeit richten. Jedoch werden gravierende Umstrukturierungen in den Verlagen die Folge sein. Und ich bin hautnah dabei. Juhu.

Wie Google Analytics nicht nur das Web erobert

Universal Analytics LogoMit dem 2013 erscheinenden Update f├╝r Google Analytics mitsamt runderneuertem Trackingcode wird sich einiges ├Ąndern. Analytics breitet die Arme (oder die verschrieenen Kraken-Tentakeln) aus und greift im Web und in der Offline Welt umsich. Was soll da nicht alles an Daten flie├čen: An der Supermarktkasse, aus dem Tel.Serviceportal und nat├╝rlich von allen digitalen Ger├Ąten (Smartphone, Tablet, Rechner). F├╝r alles kommt eine Schnittstelle in Google Analytics, oder besser gesagt Universal Analytics.┬áWie funktioniert dieses Konzept? (mehr …)