By Arvind Purushothaman
Once organizations start thinking about improved outcomes, they have to revisit the customer experience as a whole. Digital transformation is a key mantra to improve the outcome. It can be broken down into improved user experience, process transformation, and personalization using data—all supported by an underlying infrastructure transformation for improved agility.
Data—and by its extension, advanced analytics—can play a critical role in this. There are two primary challenges with data—one is collecting all of the data, and the second is understanding it.
Data Collection
If you look at the data collection from a traditional sense, the points of entry include transactional systems like enterprise resource planning, customer relationship management, other transactional systems, and custom applications. These systems are good at collecting structured data. Newer data sources, which can yield valuable insights, include logs—application logs, mobile logs, and web logs; sensors; and unstructured information from external and even internal sources. Every touchpoint with the customer should be captured. It may not be possible to define the use case for the data captured up front, but, nonetheless, this is an important first step. This is one of the reasons for creating a Hadoop-based Data Lake.
The second step is to catalog the data captured—this is an important step in the journey. Many organizations do the first step well, but often struggle when it comes to cataloging the captured data.
The third step is to assess the defined data quality boundaries—this will decide whether the data is good enough to be used for analysis purposes.
The fourth step is to come up with the transformation rules that combine data from multiple sources and create a format where it is ready for consumption. Typically, in the case of customer data, this involves building a 360-degree view of the customer where all the information about the customer is collated and made available.
Advanced Analytics
Advanced analytics involves defining the outcomes, building models, leveraging the data made available, and also creating the supporting visualization. For improved business outcomes, the customer must have a superior experience. From a data and analytics standpoint, it is primarily about understanding the customer and their specific needs, and customizing the output that will best captures that customer’s attention.
Let’s look at the banking and financial services industry, which is facing a lot of challenges from upstarts and other competitors. On the payments front, multiple disruptors have diminished the role traditionally played by banks in processing payments. If you look at the services that have been their strength in retail banking like loans, insurance, and forex, multiple players offer similar services.
I recently took an international trip and wanted to purchase a SIM card for the country I was traveling to. I contacted the mobile provider who offered the international SIM and obtained the card. My next step was to buy travel insurance and get forex services. I planned on contacting my bank for these services. However, the SIM card provider offered both services at competitive rates, and I didn’t have to deal with multiple people. I opted for both services through the mobile provider and they even bought back my unused forex at the end of the trip. All in all, it was a great customer experience.
Now, what could the banks have done differently? They have all my profile information, including my credit card and debit card transactions over several years. They can build a Customer 360 based on the data available to them in their transactional systems, and further enhance it with data from social media and other external sources. They can leverage the historical data and build descriptive, predictive, and prescriptive models using techniques like multiple or simple regression, and clustering, to determine the prospective customers and increase their chances of up sell.
To enrich historical data with behavioral traits, we bring in unstructured text data and extract features by leveraging cognitive APIs. Social media data is the key to creating a behavioral profile of the customer—moods, tastes, and choices can be gauged effectively by using Sentiment Analysis. This gives the merchant bank insights into how a user reacts and what his probable actions are going to be.
Historical data is then collated with the data available as a result of applying the cognitive APIs. Techniques such as principal component analysis help remove unwanted variables and zero in on data most relevant for model building.
The final step is to create various models using multiple regression and random forest techniques. Once a few models are available, we can evaluate the effectiveness of models by evaluating measures, like adjusted R-squared and Gini coefficient, and tie it back to actual outcomes.
The journey from raw data to insights resulting in improved business outcomes can sometimes appear to be arduous, but if done right, can lead to improved outcomes that directly impact the top and bottom lines. SW
Jul2016, Software Magazine
