Aggregate effects of advertising decisions: A complex systems look at search engine advertising via an experimental study

Yanwu Yang, Xin Li, Daniel Zeng, Bernard J. JansenInternet Res 2018 · OpenReview

TL;DR

This paper introduces EXP-SEA, a novel simulation framework for studying the complex dynamics of search engine advertising. By modeling advertisers and users as interacting agents, they reveal how factors like social media buzz (eWOM), competition levels, and strategic bidding influence market outcomes, offering valuable insights for both search engines and advertisers.

What Problem Does This Paper Address?

Search engine advertising is a massive industry, but understanding the collective behavior of advertisers and its impact on the market has been challenging. Existing research often focuses on individual advertisers or simplified auction models. However, the real-world search advertising market is a complex system where the aggregate decisions of many advertisers interacting with users and each other create emergent phenomena. Researchers lack tools to study these group dynamics and their market-level effects, especially in realistic, dynamic environments. This paper aims to fill that gap by providing a simulation framework to explore these collective behaviors and their consequences.

Methods

The core of this research is the EXP-SEA (Experimental Platform for Search Engine Advertising), a simulation framework built upon an artificial society or agent-based modeling approach. This means they create a virtual world populated by "agents" representing individual search users and advertisers.

  • Artificial Society: Search users and advertisers are modeled as agents interacting within a simulated search engine environment.
  • Cellular Automata (CA): The interactions between agents are governed by CA rules. This allows for simple local interactions to generate complex emergent behaviors at the system level. Search users and advertisers are placed on a 2D grid, and their states (e.g., search intent, budget) evolve based on their own rules and the states of their "neighbors" (other users or competitors).
  • Search User Model: Users are characterized by demographic factors (age, education, etc.) which influence their search interests and intents (informational, navigational, transactional). A key innovation is the inclusion of electronic word-of-mouth (eWOM), modeled as a "cross-over effect" between neighboring users, influencing their search intent evolution.
  • Search Advertiser Model: Advertisers have budgets, bid prices, and strategies. Their actions are triggered by user queries.
  • Search Engine Environment: This includes simulated ad retrieval, ranking, and pricing mechanisms (like Generalized Second Price - GSP).

The framework is validated through three experimental studies:

  1. eWOM Impact: Examining how social media buzz affects market outcomes.
  2. Competition Level: Investigating the effect of varying numbers of advertisers in an auction.
  3. Strategic Bidding: Analyzing the impact of advertisers using dynamic bidding strategies versus static ones.

Datasets

The paper mentions using data from CNNIC (China Internet Network Information Center) surveys for demographic statistics of Chinese search users. This data is used to inform the profiles and behavior models of the artificial search users within the EXP-SEA framework. No new benchmarks are introduced; rather, existing survey data is leveraged to create realistic agent profiles.

Key Insights and Ideas

  • eWOM Matters: Social media buzz isn't just for social platforms; it significantly boosts search engine advertising performance, increasing impressions, clicks, and ultimately, market profit. This highlights the interconnectedness of online marketing channels.
  • Competition is Good (for the Market): Increased competition among advertisers generally leads to better market performance, including higher profits. This suggests search engines have an incentive to foster competitive environments.
  • Strategic Bidding is a Double-Edged Sword: While dynamic, strategic bidding can help individual advertisers optimize their performance, there's a "cut-off point." If too many advertisers adopt aggressive dynamic strategies, it can lead to market instability and reduced overall profit. This implies a need for careful consideration of market-wide strategy adoption.
  • Complex Systems Approach is Crucial: The paper demonstrates that understanding search advertising requires looking beyond individual actions to the emergent properties of the entire system. Agent-based modeling provides a powerful lens for this.

Key Results

  • eWOM: The presence of eWOM led to a significant increase across various advertising metrics, including a higher market profit (367,207 with eWOM vs. 308,617 without).
  • Competition: Increasing competition levels (from 4 to 24 advertisers per auction) monotonically increased market profit, from 159,715 to 367,207.
  • Strategic Bidding: Market profit remained relatively stable with up to 50% strategic bidders but began to drop significantly when the proportion of strategic bidders exceeded 75%.

Limitations and Future Work

  • Advertiser Data: The current experiments rely on assumed distributions for advertiser budgets and bid prices due to a lack of detailed empirical data on advertiser preferences. Future work could incorporate richer advertiser information.
  • User Type Evolution: The study assumes search user types are fixed, while in reality, they might evolve.
  • Model Complexity: While the framework is designed for reasonable complexity, further improvements in modeling user and advertiser decision processes are planned.
  • Future Directions:
    • Enhance the modeling of search users and advertisers to better reflect their internal states and decision-making.
  • Calibrate parameters using real-world data.
  • Investigate more complex advertising strategies and benchmark against established standards.
  • Incorporate interactions with other advertising channels (e.g., TV, display ads) to create a more holistic view of the advertising ecosystem.

Summary updated

Abstract

The purpose of this paper is to model group advertising decisions, which are the collective decisions of every single advertiser within the set of advertisers who are competing in the same auction or vertical industry, and examine resulting market outcomes, via a proposed simulation framework named Experimental Platform for Search Engine Advertising (EXP-SEA) supporting experimental studies of collective behaviors in the context of search engine advertising.,The authors implement the EXP-SEA to validate the proposed simulation framework, also conduct three experimental studies on the aggregate impact of electronic word-of-mouth (eWOM), the competition level and strategic bidding behaviors. EXP-SEA supports heterogeneous participants, various auction mechanisms and also ranking and pricing algorithms.,Findings from the three experiments show that both the market profit and advertising indexes such as number of impressions and number of clicks are larger when the eWOM effect is present, meaning social media certainly has some effect on search engine advertising outcomes, the competition level has a monotonic increasing effect on the market performance, thus search engines have an incentive to encourage both the eWOM among search users and competition among advertisers, and given the market-level effect of the percentage of advertisers employing a dynamic greedy bidding strategy, there is a cut-off point for strategic bidding behaviors.,This is one of the first research works to explore collective group decisions and resulting phenomena in the complex context of search engine advertising via developing and validating a simulation framework that supports assessments of various advertising strategies and estimations of the impact of mechanisms on the search market.
BibTeX
@misc{yang2018aggregate,
  author = {Yanwu Yang and Xin Li and Daniel Zeng and Bernard J. Jansen},
  title = {{Aggregate effects of advertising decisions: A complex systems look at search engine advertising via an experimental study}},
  year = {2018},
  eprint = {2203.02200},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2203.02200},
}
Figure