PureTech Global Leverages AI to Optimize Billing Revenue

With industry deregulation and the emergence of technologies such as 5G, telecommunications companies (telcos) are increasingly looking for ways to diversify revenue streams. Mobile value-added apps and entertainment over-the-top (OTT) services are one way for telcos to inspire loyalty—and extra spending—from their subscribers.

PureTech Global develops mobile advertising solutions and content such as games, wallpaper, and video streaming for telco subscribers across 13 countries. Launched in Malaysia in 2008, the company has relied on Amazon Web Services (AWS) for its IT infrastructure since 2014.

Mindful of the disruption occurring in the telco and mobile industry and the opportunities this disruption presents, PureTech resolved to take advantage of more modern technology applications such as artificial intelligence (AI) and machine learning (ML) in the cloud. The company kicked off its first AI project in 2020, embedding AI into its native workflows to leverage consumer data and deliver more personalized, impactful apps that increase revenue for the company and its telco clients.

Without a dedicated data team or prior AI experience, PureTech engaged eCloudvalley, an AWS Premier Consulting Partner, to plan and implement the first phase of its AI initiative for its telco clients in Malaysia. The objective was to use AI to predict transaction success rates for monthly billing renewal notices.

Previously, PureTech had been sending automated payment renewal requests at the same time each day for all subscribers of its apps. However, the company speculated that if it customized renewal notices per app to send at targeted times, based on AI-generated insights of past subscription data, it could boost revenue. As an example, gaming customers might be more inclined to view messages and renew subscriptions in the evening, while customers using news apps might be more likely to do so in the morning.

eCloudvalley implemented a time-series forecasting model using Amazon Forecast to predict the optimal times to send renewal notices for each of PureTech’s apps. Engineers used two years of historical data for 300,000 telco subscribers across 10–20 apps and 4 different telco companies. After six weeks of planning and model training, the models were put into operation in September 2020.

PureTech now has a data lake built on Amazon Simple Storage Service (Amazon S3), which stores the raw historical data feeding AI models and the insights generated after forecasting. PureTech uses AWS Lambda as a serverless compute engine to provision compute resources for analysis and forecasting. Data is then sent to AWS Glue for integration. Additionally, PureTech uses Amazon DynamoDB as a low-latency database storing forecast data, and AWS AppSync develops the application program interfaces (APIs) used to automatically send renewal notices to customers.

The AI models developed entail little maintenance overhead. “We only need one engineer to manage our AI-enhanced billing renewal system, and he has no AI background. Without the help of AWS and eCloudvalley, we would have required extensive research and training, plus a whole team to handle this project,” says John Lim, chief technology officer at PureTech Global.

Lim adds, “The project would have taken at least six months, and even then the results might not have been what we wanted. eCloudvalley has facilitated the smooth implementation of a project that was completely new for us; they’ve been very responsive and have explained things in detail every time we have a question, so we can continuously improve our AI knowledge.”

Since implementing the AI model on AWS, PureTech has seen an increase of 20 percent in its billing renewal success rate. The company is using the additional revenue to design new content and deliver more mobile marketing campaigns, which are next in line for AI technology applications. The second phase of the AI project with eCloudvalley is to analyze past campaign data and derive insights that will help the marketing team design more effective, targeted campaigns in the future.

“Even in the first few months post-implementation, the AI project has generated more revenue for us. This, in turn, allows us to provide more value-added services and quality content to subscribers,” Lim says. The AI models were designed in a manner that allows for continual updates in response to fluctuations in business and market dynamics. The models in use are currently being adjusted as more data is added to the pipeline each day—and more data equals more fine-tuned results for these types of projects.

PureTech is also looking at how to utilize its newly created data lake to improve reporting. Sales and revenue reports are reviewed daily to monitor business performance. With a structured data lake on Amazon S3, PureTech benefits from more granular data access and insights. “We have so much raw data and were only touching the surface of possibilities before. Moving to a data lake on Amazon S3 enables us to analyze information on a deeper level and identify new business models,” Lim explains.

PureTech has plans to expand its international presence and is considering ways to extend its AI billing system to other countries. This is just the start of an AI journey that has the potential to transform the company’s operations across markets and lines of business.

“We see AI as a way to improve our existing workflows while we explore new opportunities,” Lim says. “After using AWS for many years, we can attest that it’s powerful and simple to use. We have a lot of support from AWS and its partners and continue to benefit from AI tools on the platform.”

The diagram below illustrates the AWS infrastructure at PureTech Global:

To learn more, visit Machine Learning on AWS.