Looking back at the beginning of the 21st century and comparing it to our current time, one cannot deny how important technology has become as an integral part of each person’s everyday life no matter what kind of lifestyle they might be leading and how fast it has developed. With its development came a surge of valuable raw material, data. In order for the data gathered to be actually useful it must undergo analysis first. This process generates knowledge and information that can be of value in many sectors. In this specific situation it was applied in marketing. Using marketing analytics would maximize effectiveness, optimize return on investment, and save both time and money used for future projects.




Key Performance Indicators

Key Performance Indicator (KPI) is a type of performance measurement that assesses how successful an organization or an activity is in achieving key business objectives and reaching targets. In this section I am going to demonstrate how to effectively track and analyze a company’s KPI. Choosing the right metric to track depends on the industry and business goals. In this case I will be using UCI’s Bank Marketing Data Set as an example, which is publicly available for research purposes.

The data is related with direct marketing campaigns of a Portuguese banking institution. The marketing campaigns were based on phone calls. Often, more than one contact to the same client was required.

Attributes:

  1. Age (numeric)

  2. Job

  3. Marital status

  4. Education

  5. default

  6. housing loan

  7. personal loan

  8. contact

  9. month

  10. day_of_week

  11. duration

  12. campaign

  13. pdays

  14. previous

  15. poutcome

  16. emp.var.rate: employment variation rate

  17. cons.price.idx: consumer price index - monthly indicator

  18. cons.conf.idx: consumer confidence index - monthly indicator

  19. euribor3m: euribor 3 month rate - daily indicator

  20. nr.employed: number of employees - quarterly indicator





Feature Engineering

We start by calculating the Conversion Rate. This is done by taking the sum of subscription and dividing it by the total number of clients.

\[ConversionRate=\frac{\sum subscriptions}{\sum clients}100\]

Exploratory Analysis

As the graph shows. It is clear that higher age groups (60 years and older) have the highest conversion rate while ages from 40-50 years have the lowest. In the second graph another variable is added. When the clients’ marital status was considered it was seen that higher ages still had the highest conversion rates. With married people who are 60 years or older having the highest conversion rates. 20-30 years old divorced clients showed very little to no conversion rates.



Conclusion

This analysis represents a short example of how KPI can be used and how different variables affect the outcome. We can further analyze the effect of many other variables including education, previous outcomes, and economic indices among other things. The information produced can then be used for a variety of different purposes.




Regression Analysis

In this section I will continue to use the previous example to demonstrate regression analysis and explain the drivers behind KPI. As it is well known,logistic regression is a predictive analysis used to describe data and explain how certain variables are related. It was used in this case to understand the variables affecting the conversion rate.

\[LogisticRegression=log\left(\frac{p(x)}{1-p(x)}\right)=\beta_0+\beta_1x_1...\beta_qx_q\]

The table provided shows the fitted model’s summary. Taking a closer look at the content will tell you that having a blue collar job has a negative effect on the outcome with high significance level as expected. Also, contacting clients by telephone showed a negative relationship in contrast with contacting clients through cellphone. Contacting clients in June, March, May, and November has a negative impact unlike contacting them in August. Students and retirees show a positive relationship. previous successful conversion showed a positive strong relationship with a coefficient of 1.32.




Predicting the Marketing Outcome

We can use logistic regression to predict which client contact will have a positive outcome. fitting the previous regression model on unseen data can predict the outcome better than random guessing with an ROC curve area of +0.86 while showing higher sensitivity for negative outcomes.







Product Analysis

Product analysis -by definition- is the process of transforming products’ data into a comprehensive and structured insight. By doing that we move from simply analyzing consumers to actually understanding why they engage the way they do with different products and services. This process gives us a very valuable perspective on consumers’ behavior. It starts with tracking and analyzing consumers’ footprints which is then used for product improvement and optimization.

I used an online retail dataset from the machine learning open source repository which contains transactions for a UK based online retail. The company mainly sells unique all-occasion gifts. Many customers of the company are wholesalers.

Attributes:

  1. InvoiceNo: A 6-digit integral number uniquely assigned to each transaction.
  2. StockCode: A 5-digit integral number uniquely assigned to each distinct product.
  3. Description: Product name.
  4. Quantity: The quantities of each product per transaction.
  5. InvoiceDate: The day and time when each transaction was generated.
  6. UnitPrice: Product price per unit in sterling.
  7. CustomerID:Integral number uniquely assigned to each customer.
  8. Country: The name of the country where each customer resides.

Exploratory Analysis

Time series Trend

When faced with similar data the first thing to do would be to form a better understanding of the overall performance of the business. As seen in the chart below, there is an upward trend which can be explained either by seasonality or growth of the business. A specific explanation can not be made because of the lack of previous years’ data.

Customers behaviors and sales

We notice from the chart below that the average customer purchases less products from April to July. This correlates with the business’s revenue. Further investigations can be done by looking at both the average unique number of consumers each month and the business’s ability to retain current customers.