How to predict Ad Viewability

Only a decade ago, the main task of marketing specialists was to find where their target audience spends most of the time. Today the answer is simple – they are all on mobile. The development of mobile marketing makes it so much easier to target the right audience, as programmatic technologies can deliver targeted digital advertising “right in the hands” of the audience.

But we are now facing a question of the next level: How to make sure that the audience not only receives advertising but also sees it?

Today advertisers are concerned that regular CPM or CTR do not allow to objectively judge the performance of ad campaign. Moreover, empty unsecured traffic leads to a lower CTR, while actual displays are more likely to give a higher score. Thus, marketers turn to Viewability measurement to be sure that their ads have been shown.

According to Viewability Benchmark Report released by Metrics, in the 4th quarter of 2019, only 53% of display advertising was visible to users by IAB and MRC standards internationally. In Switzerland, the figure was 50%, in Austria – 58%, in Germany – 46% and in France – 47%.

More and more of our clients want to assess the effectiveness of their ad campaigns as accurately as possible. Therefore, in this article, we want to talk to you about Ad Viewability that can take your business to the next level.

What is Viewability?

Ad Viewability is a metric that allows us to track the actual display of an ad unit and the time of its location in the visible area.

The fundamental standard of the Ad Viewability calculation is the IAB standard. According to IAB, an advertisement is considered as viewed if more than 50% of the banner was in the visible area for more than 1 second, and more than 50% of video ads should be visible for more than 2 seconds. Thus, advertisers will be 100% sure that their ads were shown to potential users.

Today, there are several reasons why the user may not have seen the advertisement.

Digital advertising: shown but not seen?

  • The most common reason why advertising is not visible for mobile users is when an ad unit was loaded outside the visible area of the browser. This happens if users suddenly stop scrolling the page, switch to a different one, or close the browser or app.
  • A user had left the page before the banner was completely loaded. Have you ever accidentally clicked the page and closed it immediately before it has been loaded? This happens to all of us. You might not load the page completely to see the ad, but that is enough for publishers to push the ads to the page and count it as shown.
  • Browser does not support the necessary plug-ins for ad displaying, or users use a strong firewall.
  • The page is loaded with a bot that interacts with the site unnaturally quickly. Yes, fraudulent activity is widespread on many publishers’ pages. This is an easy way for them to double or triple their revenue stream. To learn more about ad fraud, get your free copy of AD.INSIDER featuring the exclusive interview with one of the most famous independent ad fraud investigator, Dr. Fou.

All these reasons reduce the accuracy of campaign performance analysis. However, if we take the number of actual ad units displayed as the starting point, which show by Ad Viewability, advertisers will get an accurate and transparent ad performance analysis.

Measuring Mobile Ad Viewability

As it turns out, measuring Viewability in mobile is far from being a trivial topic because every environment is different: devices differ in operating system and model, and the ad itself might be placed either on a mobile website or in an app. To cope with the diversity of this situation, the ads delivered by Adello are featuring multiple viewability sensors. With those, we reach a technical Viewability measurement of 97%[1] of all cases. This number does not include the view itself – it refers only to the possibility of measuring it. Out of 100% of observed cases, in 97% we can measure.

The answer to the question of whether a view will happen or not depends on several aspects. Three examples are provided in the following:

  • Location: an ad that is placed in the header area (above the fold) of a website is more likely to be seen than an ad placed in the footer area.
  • Screen Size: a device featuring a bigger screen automatically covers a bigger part of a website, increasing the likelihood for an ad to been seen immediately after site loading.
  • Environment: in contrast to websites, many apps choose a layout where the user does not need to scroll at all. In this case, an ad is visible immediately. This is confirmed by the fact that apps yield a significantly higher viewability rate than mobile websites.

On average, a view happens in about 55% of all ad impressions delivered. However, broken down by apps and mobile websites, every value between 0% and 100% is encountered.

In summary, two uncertainties need to be considered to evaluate whether a view will happen or not: is it possible to measure Viewability, and if yes, will the user see the given ad impression at all? As explained above, both questions depend on multiple factors, and they need to be evaluated for each bid request.

Pay per view not per display

At Adello, we solve most problems with Machine Learning. For this problem we apply neural networks which are trained using recently collected data. These networks are trained with the precise goal of answering the previously formulated questions and make use of all available features that can be known prior to bidding. Being exposed to an advertising market that constantly changes, we keep our Machine Learning models up-to-date 24 hours a day, seven days a week. Our ML never sleeps.

Applying those and additional models, the probability is computed whether a view will happen or not before the actual bidding. Having this kind of information is a big advantage, and it allows us to deliver exceptional campaign results: with the so-called vCPM campaign, we guarantee our customers only pay for ad impressions that were actually viewed by their intended audience. For our clients that means: the accuracy of our pre-bid prediction ultimately lowers the cost of actually reaching the desired audience. The accurate prediction of our vCPM product is confirmed by 1.5-5x higher conversion rates compared to “conventional CPM”.


[1] all proportions are provided with respect to the total amount of inventory classified by Adello as brand-safe.

Leave a Reply

Your email address will not be published. Required fields are marked *