Optimizing direct response in Internet display advertising

https://doi.org/10.1016/j.elerap.2011.11.002Get rights and content

Abstract

Internet display advertising has grown into a multi-billion dollar a year global industry and direct response campaigns account for about three-quarters of all Internet display advertising. In such campaigns, advertisers reach out to a target audience via some form of a visual advertisement (hereinafter also called “ad”) to maximize short-term sales revenue. In this study, we formulate an advertiser’s revenue maximization problem in direct response Internet display advertisement campaigns as a mixed integer program via piecewise linear approximation of the revenue function. A novelty of our approach is that ad location and content issues are explicitly incorporated in the optimization model. Computational experiments on a large-scale actual campaign indicate that adopting the optimal media schedule can significantly increase advertising revenues without any budget changes, and reasonably sized instances of the problem can be solved within short execution times.

Highlights

► We consider an advertiser’s revenue maximization problem in direct response Internet display advertisement. ► We formulate the problem as a mixed integer program via piecewise linear approximation of the revenue function. ► Ad location and content issues are explicitly incorporated in the optimization model. ► Computational experiments on a large-scale actual campaign suggest the problem can be solved within short execution times.

Introduction

Since the first banner ad appeared on the Internet in 1994, Internet advertising has become a multi-billion dollar a year global industry; significantly surpassing radio advertisement and becoming the third largest market right behind TV and newspapers (Silverman 2010). The top two forms of Internet advertising are paid-search and display advertising. In paid-search advertising, advertisers pay an advertising fee, usually based on ad views or click-throughs, to have their websites shown in top placement on search engine result pages. In Internet display advertising (IDA), which is the subject of this study, advertisers reach out to a target Internet audience via some form of a visual advertisement such as display banner ads, flash-based rich media, or digital video. One study indicates that IDA has grown into a $17 billion global industry in 2009 (Soriano 2010), and another one reports that US. Internet users received a total of 4.9 trillion display ads in 2010 (Radwanick 2011).

Traditionally, objective of an IDA campaign is characterized as being either branding or direct response. Branding campaigns are long-term advertisement investments with goals such as boosting brand awareness, generating new customer lead, and improving customer relationship (Hollis 2005). In practice, branding campaigns aim to maximize the reach of the campaign, i.e., the proportion of the target audience exposed to at least one ad. In contrast, the goal in a direct response campaign is to achieve a measurable, direct, and immediate response. In general, direct response campaigns try to maximize revenue obtained by click-through or view-through conversions.

The objective of an IDA campaign depends on the specific business needs of the advertiser. Hollis (2005) suggests that the two paradigms, i.e., branding and direct response, are not contradictory but they are in fact complementary and that the applicability of either model depends not only on the intent of the advertiser but also the mindset of the audience. It can thus be argued that the two objectives are not mutually exclusive, yet, at a conceptual level, how the advertising budget should be prioritized is usually a matter of debate, and it is beyond the scope of this study. Nonetheless, direct response campaigns seem to be especially popular in the online arena. It is estimated that direct response advertising accounts for about 75% of all ad dollars spent online (Cox et al. 2010). One possible explanation is that technology allows for straightforward measurement of return on investment (ROI) in direct response IDA campaigns, whereas measuring ROI in branding campaigns is usually a long-term effort requiring elaborate experimental designs spanning across several weeks or months, or even years.

Despite its significant market share and vast popularity, currently there exists only one study in the literature on IDA optimization from an advertiser’s perspective. Specifically, the work by Danaher et al. (2010) is the first of its kind on optimal Internet media selection where the authors model Internet media exposure via multivariate negative binomial distribution and use nonlinear programming to maximize reach in IDA campaigns. The methodology presented therein, however, has two limitations. First, it can only be used for selection of websites. Second, it can only be utilized in optimization of branding campaigns. The purpose of this study is therefore to develop a mathematical model for optimization of direct response IDA campaigns that explicitly incorporates ad location and content issues as well as click-through and conversion rates. We also illustrate the efficiency and ease-of-use of our methodology on a $285,000 actual advertising campaign.

The mathematical model presented in this study is for maximizing revenues of a single advertiser conducting a particular direct response Internet advertising campaign. That is, the optimization problem we consider is from an advertiser’s point of view. It is important to distinguish this problem from that of Web publishers whose goal is to schedule and place ads from multiple advertisers to maximize their revenues.

The rest of this manuscript is organized as follows. Section 2 provides a literature review, and Section 3 presents our IDA model including terminology, notation, and several fundamental IDA concepts relevant to our discussion. Also included in this section is the definition of the optimization problem and its illustration using a sample campaign data. Section 4 describes our methodology in detail and presents a mathematical model for optimization of click-through conversions. Section 5 discusses view-through conversions and shows how our methodology can be adapted for tracking view-through conversions. Section 6 illustrates application of the optimization model on our sample data, and Section 7 presents summary and conclusions.

Section snippets

Literature review

There exists a vast amount of literature on various aspects of Internet advertising. A comprehensive review of previous research on this topic within the advertising community can be found in Ha (2008). Economical aspects of Internet advertising as well as its evolution over the past decade were discussed in Evans (2009). These studies considered both paid search and display advertising. Existing optimization-oriented research on display advertising can broadly be categorized into three groups:

Terminology

An advertiser is an entity (a company, government agency, educational institution, etc.) conducting an advertisement campaign for one or more of its products or services. Typically, an advertisement campaign is conducted over multiple channels such as TV, radio, newspapers, billboards, magazines, the Internet, etc. Our focus in this work is on the Internet channel. In this regard, we define a publisher as the operator of one or more websites on which the advertiser’s ads are displayed.

Formally,

Methodology

Our methodology for solving DRIP can be summarized as follows: First, we express predicted overall insert CTR as a function of the number of inserts purchased using the insert’s exposure distribution. That is, we derive an explicit expression for tio(xi). Insert CTR along with the insert CVR and CPM are then used to compute the expected revenue from the insert as a function of the ad dollars spent on the insert. The next step is the piecewise linear approximation of these (nonlinear) revenue

Tracking view-through conversions

We now briefly discuss view-through effects in IDA and describe how our model for click-through conversions can be adapted for optimizing campaigns tracking view-through conversions. Research indicates that mature markets in North America and Western Europe consistently have very low CTRs (e.g., 0.08% in the UK and 0.1% in the US), whereas emerging markets in Asia, the Middle East, and Africa have relatively higher CTRs (e.g., 0.14% in Turkey, 0.2% in Singapore, and 0.3% in Malaysia), which in

Application of the optimization model

We now illustrate application of the optimization model on our sample campaign data. We take August advertising budget as $285,000, same as the July campaign budget. This way, comparing the optimal number of conversions to that of the estimation period gives us an idea regarding how much revenue would have increased had the advertiser adopted the optimal media schedule in the estimation period.

Exposure CTR ratios for the entire campaign is shown in Table 2. Insert-level exposure CTR ratios were

Summary and conclusions

Internet display advertising is a multi-billion dollar a year industry that just keeps growing along with the rest of Internet advertising. Of the two types of Internet display advertising, direct response and branding, the former accounts for a vast majority of the ad dollars. This study, which is the first of its kind, presents a mathematical model for the revenue maximization problem in direct response Internet display advertising. The model presented is a mixed integer program that is based

Acknowledgements

The author thanks the co-editor in chief Patrick Y.K. Chau, the associate editor, and two anonymous reviewers for their thoughtful comments and suggestions. The author also thanks Lityx LLC in Baltimore, Maryland for providing the campaign data. The author is grateful to the following individuals for several insightful discussions: Lityx President Dr. Paul Maiste, Dr. Ahmet Bulut formerly with like.com (now part of Google Inc.), and Mr. Rajeev Behera with the social gaming platform playdom.com,

References (39)

  • S. Kumar et al.

    Scheduling advertisements on a web page to maximize revenue

    European Journal of Operational Research

    (2006)
  • S. Kumar et al.

    Dynamic pricing and advertising for web content providers

    European Journal of Operational Research

    (2009)
  • Achtenberg, T. Constraint integer programming. Ph.D. thesis, TU Berlin,...
  • M. Adler et al.

    Scheduling space-sharing for internet advertising

    Journal of Scheduling

    (2002)
  • F.J. Anscombe

    Sampling theory of the negative binomial and logarithmic distributions

    Biometrika

    (1950)
  • Balseiro, S., Feldman, J., Mirrokni, V., and Muthukrishnan, S. Yield optimization of display advertising with ad...
  • Beale, E. M. L., and Tomlin, J. A. Special facilities in a general mathematical programming system for nonconvex...
  • D.P. Bertsekas

    Nonlinear Programming

    (1999)
  • Bruner, R. E., and Gluck, M. Best practices for optimizing web advertising effectiveness. DoubleClick Inc. White Paper,...
  • Cole, S. Creative insights on rich media. DoubleClick Inc. Research Report, September...
  • Cox, J., Crang, D., and Vollman, A. When advertising goes digital. comScore, Inc. White Paper, October...
  • P.J. Danaher

    An approximate log-linear model for predicting magazine audiences

    Marketing Research

    (1989)
  • P.J. Danaher

    Modeling page views across multiple websites with an application to internet reach and frequency prediction

    Marketing Science

    (2007)
  • P.J. Danaher et al.

    Optimal internet media selection

    Marketing Science

    (2010)
  • A. Dickinger et al.

    Compensation models for interactive advertising

    Journal of Universal Computer Science

    (2008)
  • D.S. Evans

    The online advertising industry: economics, evolution, and privacy

    Journal of Economic Perspectives

    (2009)
  • G.M. Fulgoni et al.

    Whither the click? How online advertising works

    Journal of Advertising Research

    (2009)
  • Fulgoni, G. M., Morn, M. P., and Shaw, M. How online advertising works: whither the click in Europe. comScore, Inc....
  • Goldstein, D. G., McAfee, R.P., and Suri, S. The effects of exposure time on memory of display advertisements. In 12th...
  • Cited by (31)

    • Getting a little too personal? Positive and negative effects of personalized advertising on online multitaskers

      2022, Telematics and Informatics
      Citation Excerpt :

      Personalized advertising strategies and approaches have evolved drastically in the last couple of decades, keeping pace with advancements in digital technologies (Salonen and Karjaluoto, 2016). Among the various types of online advertising (e.g., display advertising, search engine optimization, and email advertising), display advertising has become a prevalent strategy for online advertisers due to its potential to drive both immediate revenue and brand awareness (Aksakalli, 2012). Display ads might reach customers through retargeting, which involves targeting return visitors of a website or social media profile.

    • A novel methodology for optimizing display advertising campaigns using genetic algorithms

      2018, Electronic Commerce Research and Applications
      Citation Excerpt :

      There are different types of online advertising such as sponsored search engines and display advertising. In sponsored search advertising, it considers to get the business found on search engines by using related keywords; while display advertising considers showing ads to a target audience, mainly in form of banners (Aksakalli, 2012; Pandey et al., 2017). In this paper, display advertising is in focus.

    • From institutional websites to social media and mobile applications: A usability perspective

      2019, European Research on Management and Business Economics
      Citation Excerpt :

      This literature review presents the analysis of 302 full-text scientific papers through TM to offer the current research trends in IW, SM, MA usability contexts, find patterns of information from the collected data, and translate them into valuable knowledge to uncover opportunities for further research and future applications. An IW can be defined as the sum of related web pages under a single Internet domain under the control of the company manager (Aksakallı, 2012), with the purpose of serving as a point of contact of the company or organization in the world wide web, providing relevant information. This type of website has the function of providing the basic information about an institution (https://europa.eu/european-union/about-eu_en), contacts (https://www.europarl.europa.eu/portal/en/contact), service (https://online-learning.harvard.edu/) or product (http://www.pepsico.co.uk/what-we-believe/products), among other relevant information.

    View all citing articles on Scopus
    View full text