Ai Editorial: Data infrastructure – key to balancing CX and fraud prevention

29th February, 2020

Ai Editorial: Astute infrastructure that facilitates capturing of real-time data and processing the same with minimal latency is key to setting up an apt risk assessment for legitimacy of transactions, writes Ai’s Ritesh Gupta  


A key factor in sharpening a merchant’s fraud risk assessment for transactions relates to data infrastructure and its scalability. E-commerce players need to excel in this area, and ensure all of it is streamlined so that the experience of a travel shopper isn’t hindered. It is about conducting the check for the legitimacy in a fraction of a second, so that the evaluation doesn’t adversely delay the transaction/ payment. Travel merchants must be adept at probing and investigating data in real-time to sense fraudulent transactions or any other anomalous activity.

Fraud detection specialists acknowledge challenges associated with the performance of digital assets and the significance of a scalable application.

 Some aspects that must be considered before looking at the infrastructure that support real-time fraud detection:

  • Scaling is a bi-directional process: systems must scale up to meet increased demand and scale down when demand is low, as highlighted by SecuredTouch. Count on the proficiency of microservice workloads when it comes to scaling and automating.  The team at SecuredTouch chose to scale horizontally by replicating services across the cluster, or scale vertically by adding new nodes.
  • Evaluate budget: While stepping up the performance is important, one shouldn’t forget that initiatives like switching from HDDs to SSDs, adding intermediate caches etc. also tend to increase  costs significantly, according to SecuredTouch.
  • Processing data in real-time tends to be expensive and computationally intensive, so it’s critical to ensure the fields that are required in real-time are utilized for that objective, but the others are not. Also, here we are looking a real-time or near real-time processing. Then comes the methodological choice, and for this entities have to rely on the sort of data, its volume and the database.

Key infrastructure-related areas for fraud detection

The turnaround comes from having the capability to analyze data via cloud-scale data ingestion and real-time analytics. To garner and examine a huge magnitude of transaction data calls for a vigorous database component for storage and management. Plus, a large-scale distributed computing component for running algorithms is also mandatory.

Also, from infrastructure perspective, one has to do away with managing individual servers.

Streaming data requires a data architecture that can handle rapid input and on-time output with efficient data processing. At the core of the entire exercise is to bank on a query established in advance and the objective is to alter the input stream and evaluate it based on a fraudulent-transaction algorithm. And in case there is anomaly detection, the same is conveyed to the output interface.

Some of the infrastructure-related requirements when it comes to ingestion, storage, processing, and analytics :

  • A major component is a real-time streaming platform and event ingestion offering, adept at processing a big volume of events per second.
  • Another aspect is having the ability to pulling out relevant details from data streams to identify meaningful pattern and relation.
  • Focusing on stringent security, for instance, how to deploy and manage cloud set up.

Key metrics, according to SecuredTouch, in this context are the time taken for a service to receive and respond to a request, and the time it takes to communicate with the end user.

In the whole exercise, the decisions that are related to right-sizing data, data processing method, the chosen database etc. are extremely important.


Keen on exploring fraud prevention and payment-related issues?

Check-out Ai’s conferences scheduled for 2020: