Today, we’re excited to officially announce Consensus Corporation as a new customer. Read more about how Consensus is leveraging Trifacta here, and visit our customer page to learn about all of our customer stories.
Retail data powers Consensus Corporation. The company built a platform for retailers that simplifies the complex process of “multiplex selling,” or selling subscription-based products—think selling an iPhone with an AT&T data plan or a new television with a Netflix subscription. In either example, the Connected Commerce™ platform guides retail employees through the required steps of selling bundled products to consumers, offering up the right data about each product and its offerings for more effective selling.
Stopping bad actors before they act
An effective sale means more than just a satisfied customer, but also verifying the transaction as non-fraudulent. Fraud is a top concern for retailers, especially when it comes to selling costly devices such as smartphones or televisions. For every one device lost to a fraudulent credit card, retailers must sell five more in order to recoup that loss. Consensus Corporation built a fraud detection model as part of its Revenue Cloud on the Connected Commerce™ platform in order to alert retailers of potential fraud activity before it occurs.
“If a risky person orders a new phone, our alerts give retailers the ability to stop the order and tell that person to contact customer support and upload a selfie or photo ID to verify themselves,” explains Harrison Lynch, Sr. Director of Product Development at Consensus Corporation. “Or, they might just not sell to that person, for example, if the email address matches a list of bad emails.”
Consensus Corporation routinely feeds huge amounts of retail and consumer data into its fraud detection model in order to send retailers the most up-to-date information, however they knew the faster and more accurately they could update the model, the more money they could save retailers. Even just a small percentage of recouped fraud loss could save retailers millions.
Addressing the bottleneck of fraud detection: data wrangling
Consensus Corporation had relied upon a small team of developers prepare data for its fraud model, but waiting for this team to re-write scripts with each new batch of data was a painstakingly slow process. On average, Consensus’ model development time clocked in at around six weeks, which significantly delayed the company’s response time to fraud threats with the most up-to-date information.
Time wasn’t the only concern that Consensus had with regard to data preparation, but scale and flexibility, too. When confined to developers with advanced SQL or R knowledge, it was challenging for Consensus to scale data wrangling to include those with a more robust understanding of fraud data, such as the BI or product teams. It limited these experts from applying their knowledge to the data and radically changing—or iterating upon—how the data could be transformed for the best outcome. Consensus Corporation needed data wrangling to be a faster and more widespread process across the company in order to reduce its model development time.
At the same time that Consensus Corporation identified data wrangling as a significant bottleneck, it was the midst of adopting machine learning automation platform DataRobot in order to more quickly build optimal models. Best-in-breed data wrangling technologies that could integrate with existing investments, such as DataRobot, were a must in their search.
Trifacta helps reduce model development from six weeks to one
Ultimately, Consensus Corporation invested in data wrangling technology Trifacta to alleviate its fraud model development bottleneck. The partnership between Consensus Corporation and DataRobot strengthened the value of both technologies—Trifacta accelerates the time to preparing data for the model, and DataRobot ensures that the best model is built.
With Trifacta’s intuitive interface, which guides users toward the best possible transformations with suggestions and immediate visual feedback, Consensus has expanded data wrangling to its BI and product teams. Not only are these teams applying the full context needed for the data, but are wrangling data in a fraction of the time. What once took an average of six weeks has now been reduced to one.
“The flow of using Trifacta is so intuitive,” explains Lynch. “Editing to what you want is really well done, compared to SQL or R, which is so much more difficult. Plus, our BI team is now working directly with the data instead of going back to our developers with multiple requests.”
Adopting Trifacta and its integration with DataRobot has contributed to faster fraud model development and allowed Consensus to better fight fraud, one purchase at a time.
Bigdata and data center