Eko Empowers Data-Driven Decisions with an Analytics Pipeline on AWS

For many gig-economy workers in India, access to bank accounts is sparse. These workers are typically paid in cash, work odd hours, and have little to no credit history. Eko democratizes banking and financial services by helping low- to moderate-income workers digitize their earnings.

In its 13 years of operations, Eko has served more than 70 million customers and has a merchant network of 1.5 million small and medium shops that act as cash collection and deposit centers. Historically, the company primarily offered financial services and solutions that facilitated payments and money transfers. But by 2018, Eko’s founders had evolved their vision to innovate in new verticals such as lending and insurance, which included leveraging customer data to provide greater personalization.

In 2019, the company set out to build a data and analytics team charged with creating a data pipeline within three months. The biggest challenge the team faced was the complexity of Eko’s dataset, where one database alone could contain more than 800 tables. The time required to retrieve data more than two months old could take at least 10 minutes or more. Additionally, each time project teams needed historical information from one year ago or longer, they had to submit a request to the IT team and wait for hours or even days for a response.

The data and analytics team’s first priority was consolidating Eko’s data, which had rapidly accumulated over time and was spread across multiple legacy databases. The company had been running all workloads on the cloud with another provider but was open to new vendors for its analytics platform.

The team issued a request for proposal (RFP) and chose to work with Oneture Technologies, an Amazon Web Services (AWS) Select Consulting Partner. “Nearly all the proposals we received had AWS as the foundation, and the support we got from AWS during the RFP process was a deciding factor,” explains Sheekha Verma, data science manager at Eko. “Furthermore, we appreciated how Oneture was open to listening to our ideas. The Oneture team became an extension of our own, working alongside our engineers to train them during this process.”

Eko worked with Oneture to build a data lake on the AWS Cloud that offered a unified view into diverse data sources. The company now uses Amazon EMR for big data processing and AWS Glue to prepare and load data for analysis. Amazon Athena is at the core of its analytics pipeline and is used to run serverless queries from data stored in Amazon Simple Storage Service (Amazon S3).

Today, Eko retrieves data faster than ever by running its analytics platform on AWS. For example, the company regularly runs queries on the quarterly retention rate of its end customers. This query formerly took 20 minutes to process but by using the new data pipeline, it takes just 3.8 seconds.

With near real-time data available to employees throughout the company, Eko has better visibility into its daily cash flow. “That deeper understanding of where and how the money is moving, and the velocity at different points in time, those were the insights we were missing that we can now check at any time,” Verma says.

Previously, departments across Eko had to formally submit a request to IT if they required access to data. Now, customized dashboards have been created for each department. “We’ve been able to decentralize the information-gathering process, so non-engineering teams can access and interact with data in any format they want,” Verma says.

This has led to improved decision-making and productivity gains from empowered teams. Various departments still submit requests for information from the data team, but engineers are often able to adjust the requesting department’s dashboard filters so that such requests can be fulfilled independently in the future.

To better understand its customers, Eko divided customers into groups based on their consumption patterns of the company’s services. Segmenting based on volume and frequency of transaction, along with other demographic and geographic factors, has helped Eko better serve each group’s particular needs.

Since implementing its analytics pipeline in December 2019, Eko has achieved a comprehensive level of customer segmentation. Verma says, “With our platform running on AWS, we can identify the customer segments that consistently deliver high transaction volumes and focus our marketing efforts accordingly.”

Prior to the analytics project, Eko was already looking at additional ways to attract and retain customers. Based on new insights, it is implementing a loyalty program to personally contact new customers after they use Eko’s services the first time.

Although the project and the underlying churn prediction model are still in early stages, success seems imminent. “We’re predicting at least a 10 percent increase in sales conversion among our most loyal customer segment, which contributes to 70 percent of our bottom line,” Verma says.

The agility to view and act on fluctuations in customer activity has also enabled Eko to respond quickly to market shifts and demands. For instance, in early 2020 when a data analyst detected a surge in demand for one of its non-core products, the company was able to redirect part of its marketing efforts and quickly allocate additional resources to support demand for that product. “With AWS and Oneture, Eko continues to innovate ways to better serve its customers through analytics,” Verma concludes.