For this assignment, read “The Purina Story” on page 129 and the “Let's Get Technical—Data Mining” section on page 133 of the textbook. Once you have read and reviewed this information, respond to the

For this assignment, read “The Purina Story” on page 129 and the “Let's Get Technical—Data Mining” section on page 133 of the textbook. Once you have read and reviewed this information, respond to the following questions with thorough explanations and well-supported rationale.

• Compare and contrast online research and data mining. How does each affect the business strategy of a company?
• Classify the research methods used by Purina in the case study, and determine if some type of cross-selling would benefit the company’s brand. Provide support for your conclusion.
• With the privacy concerns of today’s online consumers, do you find Amazon or Purina behaving unethically in their practices? Why, or why not?
• Analyze the privacy policies of Purina and Amazon online. What parts of their privacy policies are reassuring? Is there anything of concern?

Your case study must be at least two pages in length, not counting the title and reference pages. References should include your textbook, at least one website, and one additional credible source; therefore, you should have a minimum of three sources. All sources used must be referenced; paraphrased and quoted material must have accompanying citations and be cited per APA guidelines.

The Purina Story

Nestlé Purina PetCare Company knows with certainty that Purina websites and online advertising increase offline buying. How? Through a carefully conducted study that integrated online and offline behavioral data.

Switzerland-based Nestlé S.A. purchased the Ralston Purina Company in December 2001, gaining a full line of dog- and cat-care brands such as Friskies, Alpo, Purina Dog Chow, and Fancy Feast. The firm manages more than 30 branded websites, serving the following markets: consumers, veterinarians/veterinary schools, nutritionists/food scientists, and breeders/other enthusiasts. Nestlé started its inquiry with the following three research questions:

Are our buyers using our branded websites?

Should we invest beyond these branded websites in online advertising?

If so, where do we place that advertising?

Combining ComScore Media Metrix’s representative panel of 1.5 million internet consumers and the Knowledge Networks, Inc.’s frequent-grocery-shopper panel of 20 million households revealed 50,000 consumers belonging to both panels. Researchers created three experimental cells from survey panel members, with two of the cells receiving Purina O.N.E. banner advertising as they naturally surfed the internet: a control cell (no ads), a low-exposure test cell (1 to 5 exposures), and a high-exposure test cell (6 to 20 exposures). Banner ads were randomly sent as exposure-cell subjects viewed web pages anywhere on the internet. Next, the firm surveyed all cell members to assess the brand awareness of Purina, purchase intent, and advertising awareness. Finally, the researchers compared survey results with offline buying, as measured in the Knowledge Networks panel.

Nestlé’s marketers were very interested in the study’s findings. First, banner click-through was low (0.06 percent on average). Second, when study participants were asked, “When thinking of dog food, what brand first comes to mind?” 31 percent of both exposure-cell subjects mentioned Purina. In contrast, only 22 percent of the no-exposure subjects mentioned the brand; this result clearly showed an advertising effect. Further, 7 percent more of the subjects in the high-exposure group mentioned the brand compared with those in the low-exposure group. Next, researchers reviewed the internet panel’s website viewing habits of those who purchased Purina products and determined that home/health and living sites receive the most visits from these customers. This information helped the firm decide where to place banner ads. Among the sites frequented by Purina’s market, petsmart.com and about.com enjoyed heavy usage and were thought to be great ad buys.

(Source: “Does Online Marketing . . .” 2002)

Let's Get Technical - Data Mining 

Let's Get Technical—Data Mining

In 2012, the New York Times chronicled a case where the Target department store chain used data mining to predict which customers were pregnant. Through extensive analysis of past customer purchases over a long period of time, Target researchers found that pregnant women made a series of specific changes in their shopping habits—for example switching to unscented products and buying lots of cotton balls. Not only could the researchers predict pregnancy, but they could even predict due dates within a narrow time frame. Understanding that pregnancy is one of the few times in life that a company can garner a loyal customer, Target would then send ads for baby products to the expectant mother. So as not to spook her, Target would include ads on the same page for unrelated items—such as lawnmowers. Nonetheless, one father of a high school girl took offense and asked a confused store manager why his daughter was receiving ads for baby products. A short while later the father called back to apologize.

Data mining is the search for information hidden in large databases. It is much like scientific inquiry except that the subject of study is human-made data rather than nature. The larger the database, the greater the need for specialized data mining tools to spot patterns and relationships in the data. The tools themselves are quite sophisticated and data miners tend to have advanced degrees—with special emphasis on statistical training.

A common form of data mining familiar to online shoppers is cross sell data. Based on past sales patterns, marketers can predict which products sell well together. They then present this information to the shopper with a friendly note saying something like, “customers who bought the iXT widget also bought . . .” The same idea is behind streaming music sites that learn your preferences. They look for pattern matches with other customers that listen to your same music, and then suggest other songs that they listen to. The larger the database of customers, the more accurate the predictions are likely to be.

Here is how cross sell data mining works. Imagine that Joe buys the iXT widget and the alpha widget. As it turns out 200 other customers buy both widgets together. For each customer, the database examines what items they bought in combination. Then it counts the frequency with which each combination appears. Finally it sorts the counts from largest to smallest and presents future customers with the most likely cross sell.

Amazon has taken cross-sell analysis one step further by tying page views to purchases. For example, look at a camera on Amazon and you are likely to see the message, “What Other Items Do Customers Buy After Viewing This Item?” (emphasis added). Maybe you are on the fence between two camera models. Amazon will tell you what others on the fence ultimately purchased! They do this by counting page views of every customer and correlating those views with purchase behavior. So as you click around Amazon’s website you are creating a massive amount of data that is automatically mined.

Note that the cross-sell data is the process of fishing for relationships. We are looking at the strength of correlation between every combination of variables in the database, which is sometimes called factor analysis. However, factor analysis does not show the direction of dependence—which variable causes which. Nor does factor analysis incorporate subtle hypothesis variations to find just the right fit with the data.

But data mining can be more directed—a form of hypothesis testing. The investigator forms a hypothesis (e.g., advertising dollars are best spent on existing customers) and then uses a statistical technique such as regression to test the hypothesis. This process is the foundation of most academic research. However, it is somewhat slow and requires inventive hypothesis formation. Typically tools to carry out this analysis include SPSS and SAS.

More sophisticated still is a data mining technique called evolutionary programming. A computer program adopts the role of a scientific investigator. The program forms and tests hypotheses. When it finds a hypothesis that looks promising, it varies it slightly, forming a series of daughter hypotheses. Each of these hypotheses is tested in turn looking for the best match with the data. After churning away for a while, the program reports back to the user on which hypotheses show interesting relationships in the data.

Tasks solved by data mining include predicting, classifying, detecting relationships, and market basket analysis (the cross-sell data). The packages themselves are becoming more user-friendly, which is good news for marketers. Companies such as Megaputer may one day make sophisticated data mining commonplace.

Textbook reference - Frost, R., Fox, A. K., & Strauss, J. (2019). E-marketing (8th ed.). Routledge. https://online.vitalsource.com/#/books/9781351744843