AKEGIS: automatic keyword generation for sponsored search advertising in online retailing
Introduction
Sponsored search advertising (SSA) has emerged as a new form of Internet advertising in the last decade and is the prevailing business model for generic search engines like Google or Bing. In 2017 search revenues totaled $40.6 billion in the United States alone which represents 46% of total internet advertising revenue [26]. Consumers typically accept this kind of advertising and prefer it over other forms, such as banners, because providers deliver only ads that match the consumer's search requests. Advertisers thus spend a larger share of their advertising budgets on SSA. Advertisers usually bid at auctions for specific keywords and face in this process some complex decisions that largely affect the profitability of SSA campaigns [15]. First, advertisers need to determine the relevant keywords for which they want to place bids. Second, advertisers need to define for each keyword the page (i.e., landing page) to which search engine users will be directed when clicking on the advertiser's ad. Third, they need to set a budget for their SSA campaign. And fourth, they need to determine bids that will result in a high conversion rate and ultimately high profits or customer lifetime values.
Recent research has mainly focused on determining optimal bids [4], [36], [53], [64], budget optimization [20], [33], [42], [63], and the effects of different types of keywords on an advertiser's sponsored search performance [17], [27], [29], [38], [46]. Although researchers and practitioners provide evidence which types of keywords are rather profitable, the generation of concrete keywords is still one of the major challenges advertisers face [54]. The keywords should somehow relate to the advertised goods and they should be used in queries by consumers who are likely to click on a sponsored link and ultimately make a purchase. In recent years, a limited number of studies, e.g., [31], [55], proposed methods to automatically or semi-automatically expand an existing keyword set. All these methods need an initial keyword set and thus are not suitable to completely automatically generate sponsored search keywords. However, the automatic generation of keywords constitutes one of the most promising areas of SSA because nowadays the manual generation of typical campaigns with more than 10,000 keywords is time-consuming while the limitation to a few thousand keywords is certainly not optimal in terms of profits.
This manuscript aims to close this gap with an approach for automatic keyword generation. To come up with a theory-driven design, we review studies on consumer search and decision behavior and provide in initial empirical studies strong evidences that keywords used in on-the-store searches seem to be promising candidates for sponsored search advertising. Our approach for automatic keyword generation (AKEGIS) relies on these findings and automatically i) generates sponsored search keywords, ii) identifies landing pages, and iii) suggests keywords that should be paused for SSA. In an empirical investigation of two large-scaled online stores, we show that AKEGIS clearly outperforms manual experts (state-of-the-art approach) in generating keywords. It generates three times more profitable keywords, reduces the average cost per click and increases the online store's conversion rate.
Section snippets
Keyword generation
We briefly describe the process of sponsored search advertising from the perspective of an advertiser in this section and then review literature on generating keywords for sponsored search advertising.
AKEGIS – an automatic keyword generation approach
We propose our approach AKEGIS for automatic generation of sponsored search keywords in this section. We first review research on consumer search behavior and then provide empirical evidences that an online store's search engine log may provide a list of promising keyword candidates for SSA. In such logs the search behavior and the usage of keywords during the customer journey is stored for further analysis and optimization processes regarding the internal search quality. Usually most of the
Empirical evaluation
We implemented the idea of automatically generating sponsored search keywords based on the search log of a store's internal search in cooperation with a large-scale online store and measured several SSA performance indicators before and after the implementation. We compare these performance indicators to those of a second and similar online store from the same company to capture any effects over time (like seasonality or trends). In the following, we describe the empirical setting, explain the
Robustness checks
Difference-in-difference estimations compare two regions or periods of one group that are separated by a treatment intervention. A control group is contrasted with the treatment group to estimate the trend in the treatment group for the case in which the treatment has not been set. Differences between the two regions or periods of the treatment group might also be the result of some pre-treatment trends. We conduct robustness checks to test for possible pre-treatment trends by estimating
Discussion
In this section, we discuss the implications of our study for researchers and online store managers. Furthermore, we discuss limitations of our study and provide avenues for further research.
Conclusion
Existing research has characterized consumers who use an online store's internal search as rather goal-directed [40]. Goal-directed consumers show a higher conversion probability than exploratory consumers. Keywords entered in an online store's internal search hence mainly are from consumers with a rather high conversion probability. This manuscript introduced a novel approach (AKEGIS) that allows for automatically generating keywords for sponsored search advertising based on the keywords of an
Michael Scholz (1981) studied information systems at the Martin-Luther-University Halle/Wittenberg and received his Ph.D. in December 2009 from the University of Passau. He is now an assistant professor for information systems at the University of Passau. His research has been published in journals, such as European Journal of Operational Research (EJOR), Decision Support Systems (DSS), Journal of Statistical Software (JSS), Business & Information Systems Engineering (BISE) and in a number of
References (68)
- et al.
Mapping the customer journey: lessons learned from graph-based online attribution modeling
International Journal of Research in Marketing
(2016) - et al.
Strategic bidder behavior in sponsored search auctions
Decision support systems
(2007) - et al.
Online keyword based advertising: impact of ad impressions on own-channel and cross-channel click-through rates
Decision Support Systems
(2011) - et al.
Designing online auctions with past performance information
Decision Support Systems
(2006) Buying, searching, or browsing: differentiating between online shoppers using in-store navigational clickstream
Journal of Consumer Psychology
(2003)- et al.
Measuring consumers' willingness to pay with utility-based recommendation systems
Decision Support Systems
(2015) - et al.
Synthetic control methods for comparative case studies: estimating the effect of California's Tobacco Control Program
Journal of the American Statistical Association
(2010) - et al.
Keyword generation for search engine advertising using semantic similarity between terms
- et al.
Optimal bidding in multi-item multislot sponsored search auctions
Operations Research
(2013) - et al.
Location, location, location: an analysis of profitability of position in online advertising markets
Journal of Marketing Research
(2011)
Keyword management costs and “Broad Match” in sponsored search advertising
Marketing Science
Efficiency evaluation in search advertising
Decision Sciences
How much should we trust difference-in-difference estimates?
Quarterly Journal of Economics
Search advertising using web relevance feedback
Online expansion of rare queries for sponsored search
Zooming in on choice: how do consumers search for cameras online?
Marketing Science
Consumer search activities and the value of ad positions in sponsored search advertising
Marketing Science
Investor inattention and friday earnings announcements
Journal of Finance
Using the wisdom of the crowds for keyword generation
An empirical analysis of search engine advertising: sponsored search in electronic markets
Management Science
Prospection: experiencing the future
Science
Investigating causal relations by econometric models and cross-spectral methods
Econometrica
Search engine advertisements: the impact of advertising statements on click-through and conversion rates
Marketing Letters
The impact of sharing mechanism design on content sharing in online social networks
Information Systems Research
TV's dirty little secret: the negative effect of popular TV on online auction sales.
MIS Quarterly
Searching for experience on the web: an empirical examination of consumer behavior for search and experience goods
Journal of marketing
Deal-seeking versus brand-seeking: search behaviors and purchase propensities in sponsored search platforms
MIS Quarterly
The influence of display characteristics on visual exploratory search behavior
Journal of Consumer Research
Consumer click behavior at a search engine: the role of keyword popularity
Journal of Marketing Research
Generating query substitutions
Keyword generation for search engine advertising
Cited by (0)
Michael Scholz (1981) studied information systems at the Martin-Luther-University Halle/Wittenberg and received his Ph.D. in December 2009 from the University of Passau. He is now an assistant professor for information systems at the University of Passau. His research has been published in journals, such as European Journal of Operational Research (EJOR), Decision Support Systems (DSS), Journal of Statistical Software (JSS), Business & Information Systems Engineering (BISE) and in a number of proceedings (e.g., ICIS, ECIS).
Christoph Brenner (1981) studied business informatics at the University of Hamburg. After graduating, he worked for a start-up company and shortly afterwards founded an online marketing company. Christoph is a doctoral student at the Goethe University in Frankfurt.
Oliver Hinz (1974) studied at the TU Darmstadt Business Administration and Information Systems with main focus on Marketing, Software Engineering and Computer Graphics. After receiving his diploma (equiv. master degree) he worked several years for the Dresdner Bank as a consultant for business logic. Oliver worked as a Research Assistant at the Chair of Electronic Commerce (2004–2007) and received his Ph.D. in October 2007 from Goethe University Frankfurt. He supported the E-Finance Lab as Assistant Professor for E-Finance & Electronic Markets from 2008 to 2011 and then joined the TU Darmstadt and headed the Chair of Information Systems | Electronic Markets until September 2017. Oliver is now Full Professor of Information Systems and Information Management at Goethe University Frankfurt. His research has been published in journals like Information Systems Research (ISR), Management Information Systems Quarterly (MISQ), Journal of Marketing (JM), Journal of Management Information Systems (JMIS), Decision Support Systems (DSS), Business & Information Systems Engineering (BISE) and in a number of proceedings (e.g. ICIS, ECIS, PACIS). According to the German business journal “Handelsblatt” he belongs currently to the top researchers in the management disciplines in Germany.