X

News and Views: Drive Smart Decisions with Cloud Analytics, Machine Learning and More

  • March 25, 2019

Maximize Campaign ROI Through Response Model Using Oracle Analytics

Jignesh Mehta
Director, Analytics and Machine Learning

In this blog, Sujal Parulekar (Oracle BI Practice Manager, Data, Analytics, and AI, at Wipro) explains how Wipro's campaign response model solution is built leveraging the visualization and predictive analytics capabilities of Oracle Analytics Cloud to help marketing teams identify the right set of customers to target for their products. Wipro's DDP (Data Discovery Platform) application was designed on Oracle Analytics in collaboration with Jignesh Mehta (Oracle Partner Solutions and Product Management) team.

Organizations typically spend around 5-10 percent of their revenues for marketing activities. One of the objectives of marketing is to control marketing costs while ensuring maximum reach to customers who are likely to respond positively. Mass marketing is passé, and organizations are increasingly leaning towards running targeted campaigns. This involves identifying the right customers to target for their marketing efforts—a task that isn't easy even with all the technology that today’s marketers have at their disposal. Often a lot of the budget is wasted in targeting customers who may not be interested in the products or offers. Therefore, reaching out to the right customers has become an increasingly important aspect of the promotion planning and campaign design process, to not only save on marketing spend but also on human effort.

Why Build on Oracle Analytics?

Let's deep dive into a use case where Oracle Analytics was the tool of choice for implementation, providing a single, simplified, and easy to use platform for visualization as well as for predictive analytics. Other competitors offer a separate service for every offering, making it cumbersome to manage. This use case was developed using the self-service business intelligence (BI) capabilities that enable building great looking visualization in a matter of days. The ready to use models make the task of building predictive analytics capabilities relatively simple compared to writing complex algorithms or using languages like R or Python.

Campaign Response Model Solution

For this case in point, the solution assumes a hypothetical scenario where a banking institution is trying to arrive at the right target base to bring out a new product related to term deposits. The objective of the bank is to identify the set of customers most likely to buy a new term deposit product.

The use case relies primarily on customer demographics and past campaign performance data to arrive at a typical prospect profile of a highly receptive customer. It relies heavily on the data visualization features to showcase key demographics of a target customer profile based on responses from previously run campaigns. The objective here is to understand the profile of the customers who have responded in the past (as illustrated in Figure 1).

Oracle Data Visualization is used to build the visualizations. Simple visualizations like bar, pie, and doughnut charts, which are inherent to Oracle Analytics Cloud, are used to describe the customer profile.

Figure 1: Customer profile of customers who have responded in the past

Figure 2 analyzes the typical engagement level with these customers. The past performance data gives insights into how the customer has been engaged in the past and what was the typical campaign response at that level of engagement.

Figure 2: Historical campaign performance and prospect engagement profile

Supervised learning technique is used to predict the customer behavior in response to a campaign. The solution uses the Naïve Bayes machine learning algorithm that is prebuilt into Oracle Analytics Cloud to predict the people most likely to respond to a campaign. For the predictive analytics use case, the dataset is split—70 percent of the total dataset is used for training and 30 percent is used for testing. Customer records in the test data are ranked based on their propensity to respond positively to the campaign.

Figure 3 illustrates a typical prospect profile generated by the model that gives insights into some new customer characteristics that need to be tapped.

Figure 3: Typical prospect profile

The rankings generated by the testing dataset are evaluated against the training data to measure the effectiveness of the predictive model as illustrated by the graphs in Figure 4. More details can be found in this blog discussing an effective marketing campaign.

Figure 4: Model effectiveness graphs

The graphs show that by targeting just 24 percent of the top receptive customers, the bank can reach out to 52 percent of its customers. The lift chart is used to identify customers that are most likely to respond to campaigns in future. It gives insights into how much better the response would be by using the model versus not using a model for prediction.

Building Auto Insights with Oracle Analytics

An interesting feature of Oracle Analytics called “Explain” is the ability to generate insights based on analysis of the columns in the dataset. The beauty of this feature is that the insights can be generated at runtime with a simple click of a button. The generated insights can then be added as a separate page in the visualizations. This feature was used in the campaign response model demo to analyze the Target Customers column in the dataset and find out its correlation with other columns. A request for explanation on target customers led to the following data (explained in Figure 5).

Figure 5: Basic facts about target customers column generated through the “Add Explanation” feature

Figure 6 illustrates the key drivers of the Target Customers column generated through the "Add Explanation" feature.

Figure 6: Key drivers of “target customers” column generated through the “Add Explanation” feature

Users can pick and choose visuals to be added to the canvas as shown in Figure 7.

Figure 7: Findings from Explain “Target customers” added to the canvas

A closer look at the canvas shows that the insights thus generated are close to what has been depicted on the “Potential Target customers” visuals. For instance:

  1. Management, retired and student categories are most likely to buy term deposit

  2. Optimal contact duration is between two to three minutes and above six minutes.

What the future of 'Campaign response model solution' holds

In future, this solution will likely be integrated with a marketing application like Eloqua. This will help form a closed loop marketing analytics solution with customer demographics and campaign data flowing from Eloqua and Sales cloud to OACS, and ranked prospect information flowing back into Eloqua for running targeted campaigns.

Wipro's Data Discovery Platform (DDP) is a digital platform that uses advanced technologies to surface new perceptions, understanding, and insights around business. DDP extracts deep insights from data and uses sophisticated techniques such as visual sciences and storytelling to simplify interpretation and decision making. In this blog, Sujal has demonstrated the implementation of the campaign response model app in DDP using Oracle Analytics Cloud.

@ Sujal Parulekar, Oracle BI Practice Manager, Data, Analytics, and AI, Wipro

Sujal Parulekar has close to 18 years of experience in the BI and Analytics space. She is part of the Data, Analytics, and AI service line at Wipro. She has been working on Oracle-related technologies for more than 15 years. In her current role, as the Oracle BI practice manager, she focuses on Oracle Analytics Cloud offerings, helping customers in their transformation journey to cloud.

 

Be the first to comment

Comments ( 0 )
Please enter your name.Please provide a valid email address.Please enter a comment.CAPTCHA challenge response provided was incorrect. Please try again.Captcha