The first key to understanding how to have good data mining techniques in a Card Not Present (CNP) and subscription environment is having the correct data to use and understanding the data as it relates to your business.

Depending on your business, this data could already be readily available, or you may have to develop it. Merchants should work with their Acquirers/processor to get their authorization, settlement, refund, dispute and fraud data (Visa TC-40 and MasterCard Safe Data). If you are working with any third-party vendors such as Verifi or ethoca it can also be important to have this data available for analysis. This data can then be stored and accumulated in flat files, spreadsheets, database tables or some other storage format depending on what our company has available for resources.

Using data mining techniques to find trends is only as good as the data driving the output. Having enough historical data within your storage is also crucial so that you can have benchmarks to compare what is currently happening to your historical average. Knowing this will allow you to more easily identify trends rather than looking at data with no historical context.

Cleaning the data and preparing it for analysis

First and foremost, it is good practice to make sure that merchants are reconciling the data they receive from the acquirer to the data stored internally to ensure the data being analyzed is correct and matches what the acquirer has.

One of the most time-consuming parts of this process is formatting your different data sets into tables you can efficiently use and that are able to correlate with each other. Knowing your business is essential during this process because it will allow you to understand which data pieces are essential to have within each table. For example, using your internal CRM data to match to the data you are receiving from the processor to bring in important transaction/customer attributes like product type, sale location, client, etc.

Another important part of the data is to make sure that each set has a unique identifier that will allow you to easily correlate data from one table to the other. For example, order number, ARN’s, Visa/MasterCard transaction ID’s and authorization codes will help match customer settlement data with refund, dispute, and fraud data.

These unique identifiers will help you significantly later in the process when trying to identify opportunities or weaknesses within your business. For example, having the ability to correlate your settlement data with your refund data can tell a company which products or clients have the highest refund ratio. A merchant could identify that certain product types are selling at a high level, but customer satisfaction is low because they are asking for a refund or returning the good or service.

Identifying how to use the data to find trends

Once you have all the data, it’s time to use it. This is where knowing your business and the data will be extremely helpful to the analysis. Here are some ways data can be used for a recurring subscription merchant within the CNP space:

• Authorization rates: Since subscription businesses rely on monthly or yearly authorizations for their customers, it’s extremely important for them to understand the authorization rates, not only for customers in their early installments but also to be able to track it throughout the lifetime of their member
• In the subscription space, authorization rates are tracked by installment to ensure that customer information is captured for the initial authorization but also continues to receive approvals for the following authorizations in later installments. You can also break down authorization by authorization attempt to understand if your internal logic or if a third-party vendor you may be using is working to capture settlements that had previously been declined
• Analyzing authorization rates can be one of the best ways to increase revenue for a merchant without having to spend any additional money since you’ve already attained the customer. You just need to figure out how to get them to authorize and keep them billing
• Customer retention: Within a subscription business it’s extremely important to be able to understand how long your customers are continuing to use and pay for your services or goods. You can use your settlement data to understand by product type, sale location, client, etc., how long the average customer stays with you. You can then use this data to determine segments of your business that are more profitable and focus on growing these rather than others that could not be re-evaluated or terminated
• Determining risk of products/clients: Within a CNP subscription-based merchant, it is extremely important to be able to determine the amount of risk new/existing products or clients are going to bring to your business. Correlating your settlement data with your dispute and fraud data can help you identify segments of your business that are adding significant amounts of disputes/fraud to your ratios which could be putting you in risk of card brand compliance rules. It’s important to identify product types or clients early in the process so you do not let these continue to bill your subscriptions and compound the problem in the future. Stop the problem before it gets a start.

Building models to forecast: Machine learning

Once a merchant is extremely comfortable with their internal data to identify trends in their existing business, they can then use their historical data to forecast future scenarios. Depending on the resources you have this can be done by simply using Excel models or if you have more available to you, machine learning can be used.

As you can see, data mining can be an extremely long process from start to finish, but if a merchant puts in the time and resources, there are significant financial rewards.