Increasing the Effectiveness of Online Advertising

Kristine de Valck, Professor of Marketing - October 15th, 2009
internet buzz

Key Ideas

• Spending on online advertising cannot be optimized in the same way as with traditional media.

• For the first time, researchers have developed an optimization model that can be used to rapidly calculate the greatest exposure for an online ad campaign on a given budget.

• Generally speaking, devoting the entire advertising budget to the websites with the most traffic is not the best way to maximize a campaign’s effectiveness. 

Kristine de Valck ©HEC Paris

Professor of marketing Kristine De Valck joined HEC Paris in 2004. She has been interested in online communities of consumption since her Ph.D work at Erasmus University (the (...)

See CV

At first glance, marketing seems to be an odd territory for Laoucine Kerbache to explore. Kerbache holds a PhD in industrial engineering and is an expert on supply chain modeling and optimization. Yet, he emphasizes, “Supply chain optimization is not solely related to production. Optimization is an issue throughout the business value chain, and especially in marketing, which is the direct link to consumers.” Hence Kerbache’s interest in developing tools that help media planners target the right combination of websites for their ad campaigns.


“The dotcom bubble of the early 2000s brought companies new means of advertising,” says Kerbache. Nowadays, businesses are devoting increasingly greater shares of their advertising budgets to online campaigns. In 2007, American companies spent $21 billion on online advertising, out of a total advertising budget of approximately $300 billion. Online advertising has been growing an average of 30% a year since 2003 (this figure is nearly the same in the U.S., Europe, and Asia), and there are no signs of its slowing down. Yet companies have no models to help them make decisions about online advertising. As a result, they usually rely on habit and intuition to develop online media schedules, deciding, for instance, to advertise on the most-visited websites. Research has been conducted on traditional media since the late 1960s, but none of the associated tools show how to optimize spending on online advertising, due in part to the specific ways in which online advertising is sold (i.e., pay “per click” or “per view”).


To respond to this need, Kerbache, Peter J. Danaher, and Janghyuk Lee developed an optimization model that measures online media exposure. It includes new variables and makes it possible to derive optimum online media schedules and thus budget allocation. The originality of the approach lies in its incorporation of the following factors:

• The possibility for advertisers of sharing available space on a webpage;

• The notion of a fixed budget, which makes it possible to measure audience exposure for a predefined amount of money;

• The ability to limit the number of times a person sees the same ad, and to define a maximum number of views per person (ideally between three and ten per Internet user).


The new model makes it possible to rationally select optimal online media and to predict their effectiveness. For comparison’s sake, analyzing ten websites using traditional “complete enumeration” techniques (testing websites one by one to find the biggest audience for a given budget) takes 66 hours of computer calculations and costs $50,000, whereas the new model provides an optimal solution in … 0.4 seconds! Moreover, if advertisers want to explore a range of scenarios (i.e., limiting the number of views per person or changing the budget), they need to conduct as many tests as there are scenarios. “Between 66 hours and 0.4 seconds, the savings in time and increase in flexibility are obvious,” comments Kerbache. The researchers have also rendered obsolete the common assumption that the best way to go is to advertise on the websites with the most traffic. They explain that the most effective budget allotment varies greatly depending on the advertising scenario. Kerbache also says that diversifying media is not always the answer. The only way to maximize the effectiveness of an online advertising campaign is to use an optimization model such as this one.

Based on an interview with Laoucine Kerbache and his article “Optimal Internet Media Selection” (Marketing Science , July 2009), co-authored with Peter J. Danaher, professor at Melbourne Business School, Carlton, Australia, and Janghyuk Lee, professor at Korea University Business School, Seoul, South Korea, and the article "Networked Narratives: Understanding Word-of-Mouth Marketing in Online Communities." Journal of Marketing : March 2010, Vol. 74, No. 2, pp. 71-89.


According to Kerbache, “The model is ready to be used. It was tested and validated using real data from KoreanClick.” But how exactly can people use it to make time-and-resource-wise decisions about this part of the logistics chain? Kerbache recommends taking two distinct steps:

• Step one: statistics. Collect media metric data for the websites under consideration, organize and draw from it to determine variables for the optimization model. This is a critical step, and companies may benefit from support from agencies like KoreaClick.

• Step two: optimization within constraints. The constraints are determined by the company and its specific objectives. Optimal media selection then requires using the model and algorithm developed by the researchers. “What if” scenarios may be explored, leading to better decision-making. 


Starting from co-author Peter J. Danaher’s 2007 model for predicting online media exposure, the research team developed the “Optimal Internet Media Planning” model, a nonlinear optimization algorithm for determining the best possible attribution of advertising funds to various available websites. The model was tested and validated using a database supplied by KoreanClick, an Internet media and market research and consulting firm specialized in the measurement of Internet use, audiences, and website costs in Korea. The researchers integrated this information into their algorithm and applied the model to produce optimal media distribution plans for scenarios including a range of limitations, including a predefined budget. Their paper presents several examples that demonstrate the accuracy of their model, and includes budget-attribution recommendations for managers.