1 Keyword Targeting Optimization in Sponsored Search Advertising Combining Selection and Matching

2025-04-30 0 0 967.65KB 38 页 10玖币

侵权投诉

Keyword Targeting Optimization in Sponsored Search Advertising: Combining

Selection and Matching

Huiran Li1 and Yanwu Yang2

1School of Business Administration and Customs Affair, Shanghai Customs College, Shanghai 201204, China

2School of Management, Huazhong University of Science and Technology, Wuhan 430074, China

{lihuiran.isec, yangyanwu.isec}@gmail.com

Abstract: In sponsored search advertising (SSA), advertisers need to select keywords and

determine matching types for selected keywords simultaneously, i.e., keyword targeting. An

optimal keyword targeting strategy guarantees reaching the right population effectively. This paper

aims to address the keyword targeting problem, which is a challenging task because of the

incomplete information of historical advertising performance indices and the high uncertainty in

SSA environments. First, we construct a data distribution estimation model and apply a Markov

Chain Monte Carlo method to make inference about unobserved indices (i.e., impression and click-

through rate) over three keyword matching types (i.e., broad, phrase and exact). Second, we

formulate a stochastic keyword targeting model (BB-KSM) combining operations of keyword

selection and keyword matching to maximize the expected profit under the chance constraint of

the budget, and develop a branch-and-bound algorithm incorporating a stochastic simulation

process for our keyword targeting model. Finally, based on a realworld dataset collected from field

reports and logs of past SSA campaigns, computational experiments are conducted to evaluate the

performance of our keyword targeting strategy. Experimental results show that, (a) BB-KSM

outperforms seven baselines in terms of profit; (b) BB-KSM shows its superiority as the budget

increases, especially in situations with more keywords and keyword combinations; (c) the

proposed data distribution estimation approach can effectively address the problem of incomplete

performance indices over the three matching types and in turn significantly promotes the

performance of keyword targeting decisions. This research makes important contributions to the

SSA literature and the results offer critical insights into keyword management for SSA advertisers.

Keywords: Keyword targeting, Keyword matching, Keyword selection, Sponsored search

advertising, Stochastic programming

Huiran Li & Yanwu Yang (2022). Keyword Targeting Optimization in Sponsored Search

Advertising: Combining Selection and Matching, Electronic Commerce Research and

Applications, 56, 101209. DOI: https://doi.org/10.1016/j.elerap.2022.101209.

1. Introduction

Sponsored search advertising (SSA) has become one of the most indispensable digital media

channels. In the United States, SSA spending is projected to reach $171,641 million in 2021

(Statista, 2021; IAB, 2021). SSA is a prosperous market with three types of players, i.e., search

users, advertisers and search engines, where keywords serve as a bond tying all three together

(Yang et al., 2019). In SSA, advertisers have to select an appropriate set of keywords and determine

matching types for selected keywords simultaneously. This process is called keyword targeting

(Yang et al., 2019). Keyword targeting controls the aggressive and restrictive degree to which

consumers’ searches trigger sponsored search auctions, and helps advertisers better fit their

promoted products to search engines (Kiritchenko and Jiline, 2008)

. Well-targeted keywords will

guarantee that the right advertisements are delivered to the right consumers (Yang et al., 2017).

Therefore, it is critical for advertisers to effectively make keyword targeting decisions for their

SSA campaigns.

In the literature on SSA, plenty of research efforts have been invested in formulating keyword

selection models and developing corresponding solution algorithms (e.g., Rusmevichientong and

Williamson, 2006; Kiritchenko and Jiline, 2008; Zhang et al., 2014), analyzing branded and

competitor’s keywords (e.g., Desai et al., 2014), and examining keywords’ performance (e.g., Lu

and Yang, 2017). In another independent research stream, keyword matching has been studied

extensively in recent years from various aspects: identifying high-quality keywords (Radlinski et

al., 2008; Gupta et al., 2009; Grbovic et al., 2016), profiling advertising metrics over matching

types (e.g., Ramaboa and Fish, 2018), and bidding optimization over broad match (e.g., Singh and

Roychowdhury, 2008; Even Dar et al., 2009; Amaldoss et al., 2016). Operationally, SSA systems

require advertisers to select a set of keywords and determine how these keywords will be matched

to search queries (i.e., broad match, phrase match or exact match) at the same time (Du et al.,

2017). In effect, decisions over multiple keywords might be interdependent to each other (Yang et

al., 2019). Moreover, advertisers need to make advertising decisions in realtime due to the ever-

changing nature of SSA environments (Yang et al., 2022). Conceptually, joint optimization for

several related advertising decisions can significantly improve the performance of SSA campaigns

(Yang et al., 2012; Zhang et al., 2012; Nuara et al., 2022). Thus, from both operational and

https://support.google.com/google-ads/answer/7478529?hl=en

theoretical perspectives, it’s of necessity to address keyword selection and keyword matching

problems in an integrated way (i.e., keyword targeting), in order to help advertisers effectively

reach the targeted population. To the best of our knowledge, there is no study on keyword targeting

for search advertising campaigns. This paper aims to fill this crucial gap in the literature.

In SSA, advertisers have to face many challenges while making keyword targeting decisions.

First, advertisers have no complete performance information over the three keyword matching

types for each keyword. In practice, keyword-level historical records only contain performance

indices (e.g., impressions and click-through rate) for a certain matching type chosen in past

advertising campaigns. When making keyword selection decisions, advertisers have to take the

uniform assumption about performance indices over the three keyword matching types, which

certainly results in suboptimal solutions because advertising performance indices are

systematically different over the three keyword matching types (Ramaboa and Fish, 2018; Yang

et al., 2021a). Second, the SSA environment is highly uncertain (Yang et al., 2013; Li and Yang,

2020). In such an uncertain market, advertisers must make keyword targeting decisions prior to

the realization of values for keyword performance indices (Amaldoss et al., 2016). In addition,

advertisers mostly have limited budgets for SSA campaigns (Yang et al., 2015). That is, advertisers

need to select appropriate keywords and determine matching types with consideration of their

budget constraints.

The objective of this research is to address the keyword targeting problem in the SSA context.

First, we construct a data distribution estimation model for keyword performance indices over the

three keyword matching types. It is supposed that performance indices follow the multivariate

normal distribution over the three matching types. The Markov Chain Monte Carlo method is

applied to make inference about unobserved performance indices. Second, we formulate a

stochastic keyword targeting model (BB-KSM for short) to maximize the expected profit under

the chance constraint of the budget, and develop a branch-and-bound algorithm incorporating a

stochastic simulation process for our keyword targeting model. Finally, using a realworld dataset

collected from field reports and logs of past SSA campaign, a series of computational experiments

are conducted to evaluate the performance of our keyword targeting strategy against seven

baselines.

Experimental results show that (a) BB-KSM performs better than seven baselines in terms of

profit; (b) BB-KSM increasingly shows its superiority as the budget increases, especially in

situations with much more keywords and keyword combinations; (c) our data distribution

estimation approach can effectively address the problem of incomplete information of performance

indices over keyword matching types and in turn significantly promote the performance of

keyword targeting decisions. In keyword targeting, strategies with mixed keyword matching

enriches keyword portfolios and increases the expected profit, compared with those with a single

matching type. BB-KSM can help advertisers find more high-profit yet low-cost keywords with

appropriate matching types by searching the keyword targeting space for the global optimum. This

research contributes to the SSA literature and offers critical implications for SSA advertisers.

The remainder of this paper is structured as follows. Section 2 provides a brief literature

review. In Section 3, we build a data distribution estimation model and a stochastic keyword

targeting model in SSA, and develop algorithms for our models. In Section 4, we conduct

computational experiments and report results. Finally, we conclude this research in Section 5.

2. Related Work

In the SSA field, plentiful research efforts have been made to explore search auction mechanism

design (e.g., Huang and Kauffman, 2011; Yang et al., 2020) and search user behavior analysis

(e.g., Lo et al., 2014; Vragov et al., 2019; Lian et al. 2021), empirical analysis of performance

indices (e.g., Yang et al., 2018; Jeziorski and Moorthy, 2018; Schultz, 2020; Yang and Zhai, 2022),

and advertising decisions including bidding optimization (e.g., Küçükaydin et al., 2020; Kim et al.,

2021), budget optimization (e.g., Yang et al., 2012; Yang and Xiong, 2020; Avadhanula et al.,

2021; Yang et al., 2021b), and keyword optimization (e.g., Qiao et al., 2017; Nie et al., 2019;

Scholz et al., 2019; Song et al., 2021; Zhang et al., 2021). This study focuses on one particular

type of keyword decisions, i.e., keyword targeting, which draws from two research streams,

namely keyword selection and keyword matching.

2.1 Keyword Selection

Keyword selection is the basis for the effectiveness of SSA campaigns (Szymanski and Lininski,

2018). Researchers have addressed the keyword selection problem through semantic mapping

(Kiritchenko and Jiline, 2008; Arroyo-Caada and Gil-Lafuente, 2019; Nagpal and Petersen 2020)

and optimization techniques (Rusmevichientong and Williamson, 2006; Zhang et al., 2014; Yang

et al. 2019; Symitsi et al., 2022). Based on the feature selection paradigm, Kiritchenko and Jiline

(2008) analyzed the historical performance of individual words and phrases generated from users’

queries, selected the most promising keywords extended with highly predictive (positive and

negative) words to maximize the profit, and showed that their approach can obtain high-quality

keywords and discover specific combinations of keywords. Arroyo-Cañada and Gil-Lafuente

(2019) developed a TOPSIS (Technique for order of preference by similarity to ideal solution)-

based method sorting keywords according to their distance to the positive and negative ideal

solutions, and proved that the proposed method was effective in increasing brand awareness and

traffic volumes. Nagpal and Petersen (2020) constructed a conceptual framework to identify

profitable keywords by controlling the endogeneity of competition to measure keyword relevance.

Keyword selection decisions can also be taken as optimization problems. Rusmevichientong

and Williamson (2006) identified a profitable set of keywords by sorting keywords in the

decreasing order of profit-to-cost ratio, and formulated keyword selection as a multi-armed bandit

problem, taking into account the uncertainty of click-through rate. Under the budget constraint,

Zhang et al. (2014) took the keyword selection as a mixed integer programming problem with

objectives of maximizing the profit and the relevance of selected keywords and minimizing the

competitiveness of these keywords, presented a sequential quadratic programming solver, and

showed that their method can help increase the revenue for advertisers and search engines. With

consideration of the entire lifecycle of SSA campaigns, Yang et al. (2019) developed a multilevel

keyword optimization framework to handle various keyword decisions (e.g., keyword generation,

keyword selection and keyword assignment), and showed that the proposed framework could reach

the optimum in a steady way. Based on Markowitz portfolio theory, Symitsi et al. (2022) utilized

the risk-adjusted performance to construct diversified keyword portfolios and decide which

keywords to be selected and how much to spend on selected keywords.

Another research branch focuses on empirical analysis for keyword selection, e.g., analyzing

the competitor’s keywords (Desai et al., 2014) and keyword attributes (Lu and Yang, 2017). In

order to understand strategic benefits and costs of selecting keywords about advertisers’ own brand

names and their competitors’ brand names, Desai et al. (2014) modeled the effectiveness of SSA

campaigns depending on whether competitors’ advertisements are presented on the same results

page, and found that selecting keywords about their own brand names can preclude their

competitors from buying the same keywords, while if both advertisers and their competitors select

their brand names, a prisoner’s dilemma may be created to hurt both of their profits. Lu and Yang

(2017) regarded each keyword as a market and developed a structural model to empirically

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

1KeywordTargetingOptimizationinSponsoredSearchAdvertising:CombiningSelectionandMatchingHuiranLi1andYanwuYang21SchoolofBusinessAdministrationandCustomsAffair,ShanghaiCustomsCollege,Shanghai201204,China2SchoolofManagement,HuazhongUniversityofScienceandTechnology,Wuhan430074,China{lihuiran.isec,yangyan...

展开>> 收起<<

1 Keyword Targeting Optimization in Sponsored Search Advertising Combining Selection and Matching.pdf

共38页,预览5页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

1 Keyword Targeting Optimization in Sponsored Search Advertising Combining Selection and Matching

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: