Project topics and materials
Get Complete Project Material File(s) Now! »
Ad Viewability: Definition and Market Insights
Online advertising requires the Internet to deliver marketing messages to promote a brand to consumers, to sign up for membership or to make purchases. To do so, marketers can use many types of ads (or creatives) such as banners, videos, etc., on desktop (personal computer) and mobile environments. Different participants are involved in online advertising such as the publisher who places ads into his online content, the advertiser, who provides the ads to be displayed on the publisher’s website, and potentially many other intermediaries (ad networks, data management platforms, media agencies, etc.). With the recent development of advertising technologies (adtech), pub-lishers and advertisers manage less and less manually the ads on websites. Ads are served au-tomatically by ad servers. To measure how often impressions are delivered to Internet users, publishers, advertisers and ad servers mostly use tags, a piece of HTML or JavaScript code placed on each creative to provide a complete view of campaign delivery. The tags are usually provided by a viewability vendor.
The mission of a viewability vendor is to measure the number of served and viewed impres-sions. The number of served impressions is just the number of tagged impressions. But not all served impressions are necessarily measured by vendors because of network failures and invalid (non-human) traffic issues.9 For example, some ads can be tagged but not correctly delivered or fraudulently served to spiders and bots to manipulate legitimate ad serving. As a consequence, a second measure named the « number of measured impressions » is important to consider as it cleans up invalid traffic and non-served impressions. Finally, ads can be correctly served and measured but not seen by users for several reasons. For example, the ad can be served below the fold (i.e. outside the viewable browser space) far down at the bottom of a web page. Conse-quently, « a served ad impression can be classified as a viewable impression if the ad is contained 9One of the largest studies was conducted and published in December 2014 by the Association of National Advertisers (ANA) in the US and an online fraud detection firm, White Ops. According to the numbers, 11% of display and 23% of video impressions were bot-driven. in the viewable space of the browser window, on an in focus browser tab, based on pre-established criteria such as the percent of ad pixels within the viewable space and the length of time the ad is in the viewable space of the browser » (Media Rating Council (MRC), Viewable Ad Impression Measurement Guidelines (Dekstop), 2014). The rate of ad viewability is therefore the ratio of the number of viewable impressions over the number of measured impressions.
The pre-established criteria mentioned in the quotation above have been formally defined for different ad formats by the MRC in 2014 and 2016. A display ad is considered viewable when 50 per cent of an ad’s pixel are in view on the screen (on an in-focus browser tab on the viewable space of the browser page) for a minimum of one continuous second. This standard is valid for most banners but has been extended for large ad size banners: a viewable impression may be counted if 30 per cent are in view for a minimum of one continuous second. Regarding videos, it is required that 50 per cent of an ad’s pixel are in view on the screen and that two continuous seconds of the video are played. Finally, regarding mobile ads, the MRC has issued its first set of guidelines last April 2016 and recommends to treat smartphone (excluding apps) and desktop ads the same: 50 per cent of an ad’s pixel are in view on the screen for a minimum of one continuous second.10 Since 2012, numerous studies conducted by viewability vendors have measured the viewabil-ity of publishers’ ad inventories. All studies conclude that a significant proportion of delivered ad impressions are never visible to the end user, resulting in relatively low viewability rates. comScore has been the first viewability vendor to conduct such analysis over thousands of cam-paigns spanning a mix of global advertisers who ran their ads across a variety of publisher sites and ad networks from May 2012 through February 2013. The key finding was that 54 per cent of display ads do not have the opportunity to be seen by a consumer (comScore, Viewability Bench-marks Show Many Ads Are Not In-View but Rates Vary by Publisher, 2013). Since this first and well commented statistic, other studies have confirmed this finding even if significant increases have been observed in countries like France more recently: +7.4 points between Q4 2015 and Q1 2016, and +13.1 points in one year (Integral Ad Science report: Q1 2016 International Me-dia Quality Report). In addition, high viewability inventories are relatively rare. Quantcast for example finds that « there is a very limited supply of very high viewability inventory, with viewa-bility above 80% constitutes just two to three percent of all RTB inventory in Europe (Quantcast, Viewability: What Smart Marketers Need to Know, 2016). » In the specific case of videos, Google conducted in 2015 a study of the video advertising platforms, including Google, DoubleClick, and YouTube (Google, Are Your Video Ads Making an Impression?, 2015). He finds that 54 per cent of the videos are viewable on the web across desktop, mobile and tablets (not including YouTube).
A Theoretical Model of Online Ad Viewability
We build a model picturing a media market in which Internet users visit a website to consume content and see ads. Our model involves three types of agents: Internet users, advertisers, and a monopoly publisher.
The publisher offers an exogenous editorial content of quality q, and manages its website to attract Internet users on one side and advertisers on the other side. This is therefore a classical two-sided market in which two groups of agents interact through a platform. We assume, for the sake of simplicity, that the publisher is only financed by advertising (and not by subscription)13. Advertisers pay therefore the publisher to display ads and attract consumers that are interested in their products. Advertisers are concerned about paying for ads that are seen by users and not just served, as non-viewable ads do not promote the visibility of the products company.
A novelty of the model is that ad viewability is a publisher’s decision variable. Indeed, as reported in Section 2.2, the location of ads within webpages and a website considerably affects the viewability of ads, and a publisher, such as the recent case of the Guardian, can design its website so as to increase the viewability of ads (or not).
We analyze two situations. In the first situation, advertisers do not have a technology to measure the viewability of ads on the publisher website. In this case, the publisher cannot commit to a specific level of ad viewability, in the sense that advertisers have no technology to verify the publisher decision. Therefore, they just anticipate a global level of ad viewability. In the second situation, advertisers have a technology to measure ad viewability. Hence, the publisher can commit to a specific level of ad viewability that can be easily verified by advertisers. We can therefore compare the impact of the adoption of a technology to measure ad viewability on the demands and profits of Internet users, the publisher and advertisers. Before analyzing the two situations, we describe in more details the preferences and objec-tives of Internet users, advertisers and the publisher, as well as the timing of the game.
The publisher. To maximize its profits, the publisher chooses the number of ads a to be displayed on the website but also the level of ad viewability b. We assume a ∈ [0, a], a the highest number of possible ads displayed on the website, and b ∈ [b, b] with 0 ≤ b < b ≤ 1. We assume here that websites are constrained regarding their choices in advertising, as the lenghts of articles or the network capabitilites limit to the number of ads one website can set.
The profit function of the publisher takes the following form: Π = R(ba)N (ba), (2.1).
Welfare Analysis of Ad Viewability
We determine in this section whether the introduction of a viewability technology is profitable for the market, i.e. for Internet users, advertisers and the publisher. We calculate and compare the total welfare with and without viewability technology, denoted respectively by Wv∗ and Wnv∗. The total welfare is the sum of the surplus of Internet users (Suv∗ and Sunv∗), the surplus of advertisers (Sav∗ and Sanv∗), and the profits of the publisher (Π∗v and Π∗nv). We consider Siu∗ and Sia∗, where u∗ = 1 ∗ ∗ ∗ ∗ a∗ = N ∗ bi∗ai∗ ∗ ∗ i ∈ {v, nv} and Si γbi ai θq − γbi ai dθ and Si 0 (r(ba) − r(bi ai ))da. To keep q the analysis as simple as possible, we assume R(ba) = r(ba)ba = (1 − ba)ba and κ(ba) = ba, We provide welfare calculation in Appendix, and focus in the sequel on total welfare. We
find that total welfare is not the same depending on the presence of viewability technology. Proposition 4: Let R(ba) = r(ba)ba = (1 − ba)ba and κ(ba) = ba, which drives γ˙ = q(1−2ba) ba(2−3ba) and q ≡ γ(ba2+(b∗v a∗v +γ−2)(ba+b∗v a∗v )) :
w 2(γ−1)+ba+b∗v a∗v.
(i) Wnv∗ = Wv∗ when γ > γ˙.
(ii) Wv∗ > Wnv∗ when γ ≤ γ˙ and q > qw.
Table of contents :
1 Introduction
2 The Economics of Online Advertising Viewability
2.1 Introduction
2.2 Ad Viewability: Definition and Market Insights
2.3 Related Literature
2.4 A Theoretical Model of Online Ad Viewability
2.4.1 Model Setup
2.4.2 Equilibria
2.5 Welfare Analysis of Ad Viewability
2.6 Extensions
2.6.1 Competition
2.6.2 Ad-viewability and ad-blockers
2.7 Conclusion
2.8 Appendix
3 Targeting Advertising Preferences
3.1 Introduction
3.2 Description of the Model
3.2.1 Online Users
3.2.2 Advertisers
3.2.3 Platform
3.3 Baseline Model Without Profiling Technology
3.3.1 Stage 2
3.3.2 Stage 1
3.4 Equilibrium with Profiling Technology
3.4.1 Stage 2
3.4.2 Stage 1
3.5 Some considerations about the volume of ads in equilibrium
3.6 Winners and Losers of the Profiling Technology
3.6.1 Internet user surplus
3.6.2 Advertiser surplus
3.6.3 Profits of the platform
3.6.4 Total welfare
3.7 Strategic implications and conclusion
3.8 Appendix
4 Privacy protection and online advertising market: Opt-out impact on ad prices.
4.1 Introduction
4.2 Related literature
4.3 Methodology
4.3.1 Real Time Bidding auctions
4.3.2 Recovering ad prices using a computational process
4.3.3 Empirical Methodology
4.4 Empirical analysis
4.4.1 Modeling an opt-out choice
4.4.2 Descriptive statistics
4.4.3 Econometric analysis
4.5 Conclusion and limitation
4.6 Appendix
5 Conclusion
6 Résumé de la thèse
6.1 Introduction
6.2 Chapitre 1 : Mesure de visibilité publicitaire et marché de la publicité en ligne – une analyse économique
6.3 Chapitre 2 : Analyse économique de l’introduction d’une technologie de discrimination par la nuisance publicitaire
6.4 Chapitre 3 : Impact d’une régulation de la vie privée “opt-out”sur le marché de la publicité en ligne
6.5 Conclusion
