Abstract
Advertising options have been recently studied as a special type of guaranteed contracts in online advertising, which are an alternative sales mechanism to real-time auctions. An advertising option is a contract which gives its buyer a right but not obligation to enter into transactions to purchase page views or link clicks at one or multiple pre-specified prices in a specific future period. Different from typical guaranteed contracts, the option buyer pays a lower upfront fee but can have greater flexibility and more control of advertising. Many studies on advertising options so far have been restricted to the situations where the option payoff is determined by the underlying spot market price at a specific time point and the price evolution over time is assumed to be continuous. The former leads to a biased calculation of option payoff and the latter is invalid empirically for many online advertising slots. This paper addresses these two limitations by proposing a new advertising option pricing framework. First, the option payoff is calculated based on an average price over a specific future period. Therefore, the option becomes path-dependent. The average price is measured by the power mean, which contains several existing option payoff functions as its special cases. Second, jump-diffusion stochastic models are used to describe the movement of the underlying spot market price, which incorporate several important statistical properties including jumps and spikes, non-normality, and absence of autocorrelations. A general option pricing algorithm is obtained based on Monte Carlo simulation. In addition, an explicit pricing formula is derived for the case when the option payoff is based on the geometric mean. This pricing formula is also a generalized version of several other option pricing models discussed in related studies.
Introduction: The Volatility of Online Advertising
In online advertising, most inventory is sold via real-time auctions, which are notoriously volatile. Advertisers struggle to predict costs, and publishers face revenue uncertainty. "Advertising options" were proposed as a solution: an advertiser pays a small upfront fee for the right (but not the obligation) to buy ads at a pre-set price later. This mechanism offers a middle ground between the rigidity of traditional guaranteed contracts and the unpredictability of the spot market.
However, previous pricing models for these options had two major flaws. First, they suffered from Point-in-time Bias, calculating the option's value based on the price at a single moment, which is easily skewed by a single random bid. Second, they relied on the Continuity Assumption, assuming ad prices move smoothly like stocks. In reality, ad prices are "discontinuous"—they frequently jump or spike due to sudden changes in competition or user behavior.
Proposed Framework: Path-Dependency and Jump-Diffusion
The authors propose a new pricing framework that addresses these gaps through two main innovations: path-dependent payoffs and jump-diffusion modeling.
Path-Dependent Payoffs and the Power Mean
A key innovation of this framework is the shift from point-in-time valuation to path-dependent payoffs. As illustrated in Figure 1, the option's value is derived from the average spot market price over a specific future period $[S, T]$. This approach mitigates the impact of temporary price anomalies.
To calculate this average, the authors employ the Power Mean ($\psi$), a generalized function that can represent various types of averages depending on the parameter $\gamma$ (see Table 1). This flexibility allows the model to encompass several existing option payoff functions—such as arithmetic or geometric means—as special cases within a single "master" framework.
Jump-Diffusion Models
To model the underlying "spot market" price, the authors move beyond simple smooth curves (Geometric Brownian Motion). They use jump-diffusion processes, which combine standard continuous movement with a "jump" component (Poisson process) to account for sudden, extreme price changes. Ad impressions are "non-storable" goods similar to electricity; this leads to frequent price spikes and mean-reversion, making jump-diffusion models far more appropriate than traditional financial models.
Datasets and Stylized Facts
The study validates its models using three distinct real-world datasets:
- SSP Dataset: Real-time bidding (RTB) data from a UK Supply-Side Platform, covering over 6.6 million auctions across 31 advertising slots.
- Google UK & Google US Datasets: Sponsored search data from Google AdWords, containing daily payment prices for 106 and 141 keywords, respectively.
The authors used these datasets to identify "stylized facts"—statistical properties like heavy tails and price spikes—that prove why traditional smooth models fail for ad tech and why the jump-diffusion approach is necessary.
Key Results
- Model Accuracy: The proposed jump-diffusion model significantly outperformed traditional continuous models in capturing the actual distribution of ad prices, particularly the "heavy tails" representing extreme events.
- Revenue Impact: The analysis found that in "Bear" markets (prices trending down), selling options can significantly increase a publisher's revenue because the upfront option fees provide a "floor." Conversely, in "Bull" markets (prices trending up), options are less profitable for sellers because advertisers will exercise their right to buy at the lower, locked-in price.
- Formula Validation: For the specific case where the payoff is based on a geometric mean, the authors derived an explicit pricing formula. This formula was shown to fall within the 95% confidence interval of the much slower Monte Carlo simulations, proving its practical utility for high-speed trading environments.
Conclusion and Future Work
While robust, the model currently assumes constant volatility. Future work could incorporate "Stochastic Volatility" models to account for volatility clustering (periods of high swings followed by calm). Additionally, the model assumes the publisher has enough inventory to fulfill all options; addressing "capacity risk" and the strategic game-theoretic interactions between buyers and sellers remains an area for further research.
Summary updated