2) Clean and transform the data
The next step is fixing or removing inaccurate, corrupted, incorrectly formatted, duplicate, or incomplete data and ensuring the data is in the proper format. Wrong data and wrong data formats can lead a bank to wrong conclusions and business decisions, especially if huge quantities of big data are at play.
Multiple, interconnected channels generate vast amounts of data. There will be a mix of structured data (such as clickstream data from the web channel) and unstructured data (such as speech transcripts from support calls).
So, the bank's infrastructure should be big data ready. This means that even before a bank can apply the analytics model it's developed, it first needs to clean up and transform these raw datasets.
3) Apply the framework and link it to key performance indicators (KPIs)
Lastly, banks should apply the framework they've developed to their big data architecture and link it back to their KPIs. For example, a bank's success criteria may include the rate of digital adoption, customer experience as gauged by its Net Promoter Score, the cost to serve, or the customer churn rate (figure 3).
Additionally, the framework should synthesize the insights from structured and unstructured data to identify points of high customer effort and process inefficiencies so that the bank can address them. Analytical models can use machine learning algorithms to predict customer behavior and promote proactive customer engagement.