Investigating the effect of results ranking in sponsored search

Proceedings of the American Society for Information Science and Technology 01/2011; 48(1):1 - 10. DOI: 10.1002/meet.2011.14504801071


The objective of this research was to evaluate the effect of ad rank on the performance of sponsored search advertising campaigns. In this study, we analyzed a large log file comprised of almost 7,000,000 records spanning 33 consecutive months of a search engine marketing campaign of a major US retailer. The theoretical foundation of this research is serial position effect, including both primacy and recency effects. One way ANOVA and Tamhane's T2 tests are used in this study as our analytical methods. We examined the effect of ad rank on the critical keyword advertising metrics of clicks and conversions, which translate well to the aspect of relevance in information processing theory. Findings from our research indicate that ad rank does have significant effect on both of these online advertising metrics, although it has less direct effect on conversion rates. Primacy effect was found on clicks, indicating a general compelling performance of ads listed on top positions of the first search engine results page. Meanwhile, conversion rates across all sixteen ad positions we investigated follow a relatively stable distribution, with the exception of the first two ads. Ads in these two positions generated extremely high conversion values. However, examining both clicks and conversion rates combined into a single metric of conversion potential, it is apparent that ad rank has a significant effect on the performance of keyword advertising campaigns. Ad appears at the top of the results listing have a much greater conversion potential. The research results reported in this paper are beneficial to companies using search engine marketing as they strive to design more effective advertising campaigns. Implications from this study could lead to more targeted bidding strategies and thus reduce ineffective bidding in keyword auctions.

5 Reads
  • [Show abstract] [Hide abstract]
    ABSTRACT: In this paper we consider the serial position curve in immediate verbal free recall. A large literature has argued that two components of the serial position curve, recency and primacy, reflect the functioning respectively of short-term and of long-term memory. However, there are a number of difficulties in interpreting the recency effect as a phenomenon uniquely associated with short-term memory. Moreover, the serial position curve has been used widely for clinical investigations in patients with memory deficits. This is despite the lack of norms for the measures derived from the curve. We present a set of standardised norms based on 321 Italian normal subjects. These norms are shown to be applicable both to an English speaking population, and to three groups of brain damaged-patients, namely Alzheimer's, amnesics, and frontals. The standardised norms offer a clinical and experimental tool which, coupled with a multiple single case approach, allows us to show dissociations and double dissociations among the performance patterns obtained from all three pathological groups. The paper concludes with a discussion of a possible interpretation of the recency effect as a emergent property of all types of memory system, including verbal short-term memory. Taking into account previous literature as well as our own data, the recency effect in immediate verbal free recall is here interpreted in terms of a two-component view of verbal short-term memory.
    Cortex 10/1992; 28(3):315-42. DOI:10.1016/S0010-9452(13)80143-8 · 5.13 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: As with any application of machine learning, web search ranking requires labeled data. The labels usually come in the form of relevance assessments made by editors. Click logs can also provide an important source of implicit feedback and can be used as a cheap proxy for editorial labels. The main difficulty however comes from the so called position bias - urls appearing in lower positions are less likely to be clicked even if they are relevant. In this paper, we propose a Dynamic Bayesian Network which aims at providing us with unbiased estimation of the relevance from the click logs. Experiments show that the proposed click model outperforms other existing click models in predicting both click-through rate and relevance.
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: We present a very brief history of the origins of sponsored search, which is the presentation of text advertisements in response to a user's search query.
    Bulletin of the American Society for Information Science and Technology 12/2006; 32(2):12 - 13. DOI:10.1002/bult.1720320206
Show more