Data Science for Advertisers

Data Science is very important for many everyday jobs and activities in our world, including advertising. Lots of businesses use data science to create efficient and cost effective targeted advertisements. Targeted advertisements are specific advertisements that are directed exactly at what the customer wants. Let’s start from the beginning. Have you ever been shopping for some clothes and then you go to a different website and there is an advertisements for the exact shirt you were looking at? Well that is all because of data science. Advertisement companies use things such as a user’s clickstream data to narrow down on what the customer likes. Clickstream data refers to a record of all the websites a user has viewed, such as a search history. This information is stored in a file called a digital cookie. You have probably seen a website ask if they are allowed to use cookies, and you can easily say no if you want, but most people don’t. For example, if a user’s clickstream includes golf players and golf highlights, the algorithm deduces that this person likes golf, and they may see advertisements for golfing clubs even on unrelated sites, such as the weather. Data Science is involved in this because the algorithm goes through the 4 steps. Data Collecting is done when a user’s clickstream data is collected. Data Wrangling is performed when outliers to the data set are found. For example, someone else may have used a user’s computer and searched for soccer movies, even though this user likes golf. Outliers may be things that are only searched a couple of times, and have no relation to the majority of searches. Next comes Data Analysis, when the data is examined by the advertisement companies, and they make conclusions. If someone was looking at this search history: [flights to Chicago, Christiano Ronaldo, hotels in Chicago, things to do in Chicago, definition of imperialism, Chicago clubs game times], they might make some conclusions that the user is planning to go to Chicago. Although there are 2 random searches in the mix, these are outliers to the data set. Finally comes Data Representation. In this step, the actual companies put out their advertisements. Following the previous example, these may include deals on hotels in Chicago or tickets for Chicago clubs’ games. As can be seen, targeted advertisements are better for the user because they see things that they are actually interested in, but it is also better for the company, because they advertise to the right crowd, and don’t waste money advertising golf clubs to people who don’t like sports. There are still some flaws with targeted advertisements, such as skewed data. For example, a husband could be searching for a dress for his wife, and then getting advertisements for dresses everywhere he looks even though he already bought one, but over time the algorithm eventually changes, and realizes that this customer does not want more dresses. So the next time you shop for a laptop from Best Buy and then magically see an advertisement for the same laptop you were looking at on the Macy’s website, just know that it’s not a coincidence.