Every minute, more than 3 million queries are submitted via Google’s search engine. Analyzing data on these queries, along with page results and subsequent clicks, can shed light on the kinds of information that consumers seek each day. However, search intentions may be influenced by variables such as the weather and even the TV shows aired on a certain day, according to HKUST’s Jia Liu and colleagues. Focusing on televised TV shows such as football games, they offer insights into the preferences and behavior of TV viewers that may help marketers to tailor their content to achieve highly ranked search results.
In today’s Information Age, the huge number of search engine users and the diversity of their searches make it extremely difficult to map out online search patterns—especially as ever-changing variables such as the time of day and major (sporting) events can affect what people look for online. For example, the authors explored how online search patterns changed before, during and after the airing of the 2016 Super Bowl, a major televised sporting event in the US.
“Some queries, such as ‘super bowl’ and ‘carolina panthers,’” the researchers report, “were searched for consistently over time.” In other cases, users’ intentions varied according to when they went online. Before the game, viewers searched for information on “kick-off time”; afterward, they were keen to watch “highlights” online. These findings were obtained simply by measuring changes in search volume over time. More subtle variations in content preferences were identified “by observing downstream behavior such as clicks on returned links.”
Search engines and advertisers may waste their money if their Internet advertising is not effectively targeted. This relies on being able to predict the click-through rate (CTR), or the likelihood that a user will click on a link when it appears in their search results. “It is critical for them to improve predictions for new links quickly,” say the researchers, “so they can adjust ranking or search campaign strategies.”
To address this problem, the authors developed a novel topic modeling approach to link data on searches, results and CTR. Predictive frameworks of this kind, which are commonly applied in the fields of marketing and information retrieval, can reveal patterns in huge volumes of big data. The researchers tested their approach with real-life data on US TV shows from the Bing search engine. As intended, their framework enabled them to “interpret and explain search volume, clicks on links, and the relationship between the two.”
“Our proposed approach can simultaneously capture multiple types of information and investigate multiple aspects of behavioral dynamics in a single, robust modeling framework,” the researchers conclude. This will enable valuable predictions to be drawn about future Internet user activity, adding a degree of certainty to the process of advertising online. “It can help advertisers decide which ad copy should be shown to which users in which context,” say the researchers, “and launch and optimize search advertising campaigns triggered by offline events such as TV commercials.”
By promoting targeted advertising, this innovative modeling approach will benefit not only businesses but also today’s sophisticated Internet users, who want to access the right information as quickly as possible.