Marketing Analytics & Data-Driven Solutions

Medium, Yueh-Han Chen 

Topics: Campaigns, Data, Marketing, Solutions

Transcript Excerpt
Presented by Medium is a data set containing 28 columns and 2240 rows, with 11 columns containing personal information about the customer. For instance, the customer’s identification number, birth year, educational level, location, income, and the number of minors living in their home. Additionally, eleven data columns detail the behavior of customers. For instance, the amount spent on wine over the last few years, as well as fruits and meat products, fish products, sweet products, gold products, etc. Finally, six columns detail customer responses to previous monthly campaigns. For instance, if a customer accepted an offer in the third campaign, the value of accepted campaign 3 is 1. 
Yueh-Han summarizes five data-driven solutions and provides an explanation of the rationale for each solution. To begin, he recommends repurposing the previous campaign’s techniques, but with an emphasis on meat and wine promotion. The second option is to increase marketing spending in Spain while decreasing it in India. Thirdly, hold a Thursday brand discount day or a June brand discount month. Fourth, implement marketing promotions to convert customers who primarily shop online or through catalogs to in-store shoppers. Finally, but certainly not least, a loyalty program for high-income customers should be developed. 
Yueh-Han explains each solution in detail. To the first solution, the most recent campaigns outperformed the previous campaign by nearly double. Additionally, he plots the differences in customer demographics and purchase behaviors between the most recent campaign and all previous campaigns. Customers who participated in the previous campaign spent nearly twice as much on meat and wine products as those who participated in the previous campaign. He could determine whether the total amount spent on each of the product categories, which included wine, meat, gold, fish, fruits, and sweet products, was greater than the sum of the individual product categories. Data correlation exists between the acceptance of marketing campaign offers and the effectiveness of the marketing campaign. Although they are all statistically significant, they do correlate with total acceptance of the wanting campaign, as their value is very close to zero. Because wine and meat products have a higher Pearson correlation, he suggests that the marketing campaign should continue using the same techniques as the previous campaign, but with an emphasis on meat and wine promotion. 
Moving on to the rationale for solution number two, the previous campaign resulted in a greater number of valuable customers. Yueh-Han graphs the country percentage differences between the most recent campaign and all previous campaigns. As can be seen, Spain is 4 percent higher proportion of customers than India, while India is 3 percent lower proportion of customers than Spain. As a result, Yueh-Han recommends that the marketing team invest more in Spain and less in India. The rationale behind solution three is to determine the month and day in which each customer became a customer and then use that information to create marketing campaigns on that day or in that month to attract additional customers. The rationale for solution four is that he conducted a correlation test between the average order value and the total number of indoor purchases and discovered that they are statistically significant. As a result, he recommends that the marketing team implement a marketing campaign aimed at converting customers who primarily shop online or in-store. Finally, he plots the heatmap to determine the data correlation between income, total spending, the total number of purchases, and average order value. Indeed, their correlation is strong, which is why he recommends developing a loyalty program to retain high-income customers. 


How Kismet Works

The Transparency of Product Life Cycle Data