Project Retina
At Airtel, we continuously emphasise on the importance of customer experience and improving it. Understanding customer needs and going deep into the customer problems is embedded in our DNA. This customer obsession has propelled us to deliver high-quality platforms, meeting our customers’ diverse needs. One among such customer problems was providing one view of the transaction history to all the self-care as well as the assisted channels. Thus, Retina came into inception. In this blog post, we will delve deep into Retina’s architecture and provide a detailed explanation of its design.
Problem statement
There was the need of a low latency even aggregation system which can co-relate the data from various systems and add context of received data with customer profile/hierarchy/relationships even though the data ownership lies with the respective source system/application and provides a single Payment/Refund/Waiver history view by keeping all these events aggregated at one place.
Tech solution
Retina developed the capability to provide a unified view of the customer transaction for the last 6 months by aggregating the information from multiple CRMs, payment and provisioning stacks.
Retina as a platform is processing 5 MN+ events from different systems daily (with the capacity for much more) and offering a view of the transaction status by stitching the payments, postings, refunds, reversals into a single transaction.
Low latency APIs with pagination support were provided on the top of this view to enable self-serve and assisted channels to query real time transaction status and show the customer within 1 second after the customer has initiated payment.
To ensure data accuracy, near real time recons were built to reconcile the real time data with the respective masters.
Retina as a platform is divided into 2 parts:
· Ingestion — This part is responsible for reading the events published by different systems from a queue, cleansing, de-duplicating and enriching these events and then publishing the enriched events into another queue along with staging the source events in a persistent store.
· Cache — This part is responsible for reading the enriched events, stitching the payments, posting, refund events together and storing the same in a low latency cache. APIs are provided on top of this low latency cache along with pagination capability.
Benefits
Retina is a key platform instrumental in powering the transaction history feature on Thanks App. Retina as a platform has been instrumental in providing a much more enriched view viz. reversals, refunds and various other customer transactions. Retina has really helped in delivering great experience and hence reduction in interactions and repeat.
Some of the key benefits of this platform are:
· Reliable and real time status updates- Ability to see real time (99.99%) status updates of transaction, thereby increasing customer satisfaction.
· Richer and actionable info- Ability to see detailed transaction view (transaction ID, pay mode, time etc).
· Refund/Reversal visibility- Ability to view the refund status of transaction and raise a complaint directly.
· Airtel black and multi-Telco payments- Detailed break up of payments for high value customers will now be visible.
· 40% reduction in customer interactions (SRs & FTRs) related to payments leading to better High Value customer experience
Conclusion
Post Retina integration with IVR and Thanks app, we have seen a more than 40% of call drop for transaction related queries. Retina is a powerful platform for managing the different lifecycles of a single transaction. It has ensured the standardisation of architecture across 7+ Airtel systems and ensured a consistent view of the transactions across Airtel. Retina not only facilitates centralised access but also ensures that all the respective masters are not responsible for building their localised caches to serve low latency API requests.
