Dynamic Design for Routing Customer Calls

How Capital One introduced engineering excellence in our cloud-based contact center ecosystem


Introduction

Capital One was the first US bank to declare an all-in on the cloud strategy and our cloud migration journey was successfully completed in 2020. This journey touched on every line of business, every product, and every team at Capital One. For our credit card line of business, the engineering teams played a critical role in migrating our contact centers to the cloud. These centers serve 60 million credit card customers, with over 25K agents spread across the world working 24*7/ 365 days a year servicing our customers needs.

As the tech lead for the engineering teams at Capital One Card Tech, I want to share how we transformed the Amazon Connect ecosystem by leveraging cloud computing tools such as Lambda functions, DynamoDB, and REST APIs to externalize attributes and run business rules dynamically against each incoming customer call. This article will also cover how a dynamic design in the routing world helped in improving time to market for new features and how we optimized run the engine (RTE) costs and built monitoring dashboards.

Time for a new contact center solution

In 2019 Capital One credit card agents migrated from a legacy on-prem contact center solution to Amazon Connect, which is Amazon’s cloud-based contact center solution. Migration to Amazon Connect provided Capital One with some immediate wins with respect to scalability, costs (per second usage instead of per seat) and ability to react to dynamic network changes.

As covered in our Amazon case study, the flexibility of Amazon Connect lets us add new features in weeks, instead of the three to six months that our on-prem solution required. Given the highly regulated business environment we operate in, and with the high volume of calls and changes in the network, Capital One identified some additional opportunities to streamline the contact center operations and management post migration.

Continuous improvement opportunities

Amazon Connect is primarily leveraged for routing customer calls to the right agent at any given point of time. Every call is evaluated against set parameters, and with various suites of products being offered to credit card customers, Capital One has a complex set of classification and targeting rules for routing calls. With hundreds of queues to manage, and with all the attributes being hardcoded in the contact flows, the engineering team’s primary focus post migration was to update the contact flow attributes for each of the queues.

However, we had limited capacity to innovate. Given the volume of calls coming into the contact centers on a daily basis, these changes had to be done during off business hours. Additionally, with the lack of APIs on the Connect platform in 2019, the team was limited to manual deployments taking up several hours of manual configurations in both Active & DR regions.

Building engineering excellence

The engineering team supporting our call routing work identified opportunities to streamline our cloud-based contact center solution. The team decided to externalize the contact flow and queue attributes to a database and leverage the power of AWS serverless lambda functions in calling the dynamic attributes and executing the business rules in real time every time a customer call is routed to an agent. By combining the output from Amazon’s real time metrics APIs, clubbed with the output of the dynamic design leveraged in classifying and targeting the customer calls, we were able to streamline critical functions of our contact center operations. This dynamic design leveraged our enterprise CICD pipeline for deployments, meaning any change could be deployed during US business hours without impacting contact center operations.

When Capital One completed the acquisition of BlueTarp - a leading business-to-business trade credit financing company -in 2020 one key activity was setting up their contact center in Amazon Connect and onboarding their 150 agents. With the lift the engineering team achieved with dynamic design this set-up and onboarding was completed in just 72 hours. The same effort would have taken at least three weeks if the team had to hard code all the attributes in all the non prod and production regions across regions.

Image: Dynamic design for Call routing in Amazon Connect

Image: Dynamic design for Call routing in Amazon Connect

Continuous improvements with product innovation

With the rollout of the dynamic design, the card engineering team was able to focus on building some innovative solutions leveraging the Amazon Connect platform. During the peak of the COVID-19 pandemic in 2020, contact centers across a wide range of industries saw their call volumes skyrocket. For a period of time during the initial peak of the pandemic, Capital One contact centers saw high call volumes with long wait times for customers to interact with agents.

The engineering teams leveraged Amazon’s Queued Call Back functionality and integrated it with our dynamic design. This allowed customers to retain their position in the queue without having to stay on the line; receiving a call back from a Capital One agent once their turn was up. This queued call back capability had a direct positive impact on customers by removing the need to stay on the line while providing the assurance of a call back from Capital One. It also improved the contact center metrics for the longest call in queue across the network.

We rolled out the proof of concept in 24 hrs and rolled out the solution to the entire 600+ queues in approximately four weeks. The queued call back capability not only helped in providing relief to our contact centers, but also helped our customers avoid sitting on the phone for extended periods of time while waiting for their turn. When customers took post call surveys, there was clear evidence that customers had positive feelings about the call back experience, with Capital One living up to the promises made during the call back offer. From an operations perspective, queued call back functionality helped in reducing the customer hangup rate by more than 25%. It also helped optimize costs incurred from the telephony suppliers due to customers waiting in the queues to talk to an agent.

Example feedback from customers & agents on queued call back functionality

Example feedback from customers & agents on queued call back functionality

Optimizing run the engine costs

To further optimize the Run The Engine (RTE) costs incurred in maintaining the Amazon Connect platform, the engineering team came up with an operational portal which was integrated with enterprise SSO and with the database used for storing the dynamic attributes. With the launch of the operational portal, all the operational activities were transferred to the operations team as self service capabilities, with the engineering team primarily focused on rolling out new innovations on the platform. The rollout of this dynamic design helped reduce our contact flows by 87% which helped in optimizing the RTE costs associated with the platform.

With externalizing the contact flow and queues attributes to an external database and leveraging lambda functions and APIs in calling the same, the engineering team has access to a wealth of new data in the form of logs from the cloud components. With access to the logs, we are able to build real time monitoring dashboards and integrate with enterprise alerting systems to alert us whenever a benchmark is missed. The monitoring dashboard also acts as the “single pane of glass” for all stakeholders, this ensures that everyone looks at the same (accurate) data and interprets the issue in the same manner.

Conclusion

Amazon Connect makes it easier for Capital One to not only scale on demand at an optimum cost but also provides critical customer capabilities. Implementing the dynamic design took the whole contact center technology experience - for both customers and agents - to the next level and allowed the engineering team to leverage the power of cloud components, the logs which are generated, and the dashboards to optimize RTE costs and improved time to market for delivering innovative product solutions to our customers.

 

Computer photo created by diana.grytsku - www.freepik.com


Ram Mepperla, Lead Software Engineer

Ram has been a Technology leader with over 20 years of experience in various facets of Technology & Product development. He has worked in a wide variety of industries: Big 5. Consulting, Insurance, Pharma, Investment Banking, Telecom, Banking and e-commerce. With his experience working in several continents in Europe, North America & South Asia, Ram was able gain the inclusive culture and evolved as a Belonging and Inclusive champion. His thought leadership and Strategic thinking helped in solving several complex engineering and business problems across various industries. Ram enjoys mentoring others and helps in making other successful in their respective areas of interests. He holds a BS, MBA in Marketing and an MS in Technology Management. Ram lives in Richmond, Virginia with his wife and children.


DISCLOSURE STATEMENT: © 2021 Capital One. Opinions are those of the individual author. Unless noted otherwise in this post, Capital One is not affiliated with, nor endorsed by, any of the companies mentioned. All trademarks and other intellectual property used or displayed are property of their respective owners.

Related Content