How machine learning is helping Credit Karma reintroduce itself to users

The personal finance company is hoping a new ML-based service will shift the perception of Credit Karma beyond that of just a credit score app.
Written by Natalie Gagliordi, Contributor

Personal finance company Credit Karma has always relied on the use of consumer credit data to power its services and fuel its business model.

But it's only recently that the company is turning to machine learning to make sense of hundreds of billions of data points and deliver personalized insights and recommendations to individual members at scale.

When it launched in 2008 at the height of the financial crisis, Credit Karma's primary service leveraged credit report data to help consumers understand, track, and improve their credit scores. The company managed to gain traction as a provider of simulated credit information despite a wave of consumer skepticism stemming from credit monitoring scams, amassing over one million US users in less than two years. 

Credit Karma is now hoping to reintroduce itself to its 100 million US members through a new ML-based service that's meant to deliver personalized insights and recommendations tied to the company's range of auxiliary services.

The new product feature called "Stories" is powered by an internally built technical infrastructure that leverages ML and data science to show members relevant offers tailored to their financial situation. 

The goal with Stories, and the company's use of machine learning overall, is to shift the perception of Credit Karma beyond that of just a credit score app. 

"Most people don't know that we offer all these other products," said Andreas Gross, a senior product director for Credit Karma. "We have 2,500 data points on our members collected throughout the year. Previously the data wasn't used to gain other insights, but with Stories the idea is to start customizing what you see." 

The Stories feature utilizes data in a range of ways. For instance the data monitoringcategorytells a member how they are doing on an aspect of their financial life, and the recommendation category uses personal data and insights to suggest specific actions. Meanwhile, the feature discovery helps members discover parts of the product that might be relevant to them, and the signal story asks a member a question or otherwise gets their input in order to further provide insights and recommendations.

"The best way to think about it is that we get data from credit reports and try to make the data more granular and useful," said Gross. "But we never really had this place where we could have a conversation with our members, so it's part of our future plan to get more data right from users." 

Credit Karma tested Stories with an initial roll out to around 350,000 members and saw a 3.1% increase in member engagement with the Credit Karma app. Based on Credit Karma's active iOS monthly visits, the company expects Stories to translate to an additional 1.8 million iOS visits per month once the update is fully deployed. 

"The way we look at this going forward, is that in the past we have struggled to make clear to our members that we offer these other products," said Gross. "In 2020 the goal is to scale Stories and make it a key driver as a company." 

In terms of privacy and security, Credit Karma said its privacy policy provides a detailed look into what data it collects and how it's utilized. The company said it always asks for consent to gather and use member data, whether that be in-product consent or through its terms of service. Credit Karma also insists that it never sells member data, and that it only shares data in limited circumstances with a member's permission. Security is also a major factor in the company's infrastructure strategy.

"We take platform and data security very seriously and the trust of our members is what Credit Karma is built on," said a Credit Karma spokesperson. "One of the primary reasons why we made the transition to Google Cloud is because it offers advances in infrastructure security to match our standards in keeping our big data safe. We have a security model that has been built upon over more than a decade with important features like data encryption."

Credit Karma said its transition to Google Cloud has taken place over the past two years, starting with its production site and continuing with the use of Google Cloud's data infrastructure and its suite of managed services for developers. The company said its cloud transition is an ongoing process as its back-end technology is constantly evolving.

Nonetheless, the move to the cloud has allowed Credit Karma to operate a machine learning environment that runs over eight billion predictive models daily. 

"Our predictive, data-driven recommendation systems are what uniquely positions us to help people navigate the complex landscape of personal finance," the company said. 

Editorial standards