Real time is the buzzword in payments. But how can real-time data, real-time tracking and real-time interaction with the customer actually lead to better business results?
In a webinar earlier this month that asked the question “How Can Real-Time Decisions Improve the Customer Experience?” PYMNTS’ Karen Webster noted that in order to ensure the best results for both customers and the firms that serve them, data must be good, fraud and attempted fraud must be tracked in real time, and platforms must remain robust. In discussion with a trio of Enova executives — Joe DeCosmo, Chief Analytics Officer; Mike Failor, Head of Portfolio Analytics, and Steve Lake with Strategy and Operations — a few key tenets emerged tied to best practices in building a truly responsive and interactive customer experience.
There are a number of “business blockers” in place that actually hinder customer experiences, said DeCosmo, including the lack of real-time automation in decision-making and a lack of technical platforms, in addition to the pressures of having an effective staff in place that can deal with the analytics effectively and in real time, too.
According to DeCosmo, the basic definition of a real-time environment boils down to “sub second responses” across a number of considerations, including credit risk and marketing efforts, resulting in a “faster experience … that is also more streamlined.” But in order to have that streamlined process in place, the right data model must be in place. The fact that the model itself can help make the business decisions, said DeCosmo, means less manual processes or contact people have to get involved, which can introduce more risk into the process. Having strong models in place also allow for verification decisions to be done on the fly.
And in reference to fraud and customer monitoring, said DeCosmo, focus on fraud, real-time identity and account documentation – including the ability to “wait” on debits from accounts in the event there may be an overdraft – helps improve the experience overall, without turning away legitimate transactions.
However, noted the executive “there is no perfect model” which means that firms seeking to optimize the customer experience need not spend needless time and effort and money on seeking a perfect system. The biggest question is whether to buy or build a platform, or construct a hybrid of those two options. Analytics providers can run the gamut between consultants or separately contracted analytics providers.
In discussion of its own real-time analytics platform, Colossus, DeCosmo noted that the platform uses regression and machine learning, while Failor delved into the mechanics of the Smart ACH solution that helps boost firm profitability by maximizing the debit process – as mentioned above, canceling transactions that could be negated due to insufficient funds – and which, as the executive noted, help decrease return rates. Failor noted that the construction of predictive models continues long after the initial buildout, with constant refinement of data, iterative processes and algorithms that sort through that data.
In a case study offered up by Enova’s Lake, who led the implementation for Enova’s CashNetUSA brand, which “started when NACHA announced ACH rule changes that would take effect in 2015,” with the intent to gain insight on how those changes, related to business payments, would impact then-current Enova processes. CashNetUSA was able to save more than $8 million annually in the wake of NACHA rules that helped mandate faster (and same day payments) with an initial impact on the bottom line of as much as $18 million annually.
The key was to address payment by payment decisions rather than just adopting blanket rules. “That brought us to step one: build a model based on past historical data,” said Lake, in order to predict likelihood that payments will clear. The result has been the Smart ACH solution, with the ability marry a model and business rules in tandem with examining payments on a micro level in an automated fashion. Then they rolled out a full solution in December of 2015. The eventual refinement of the model, and a new iteration, brought in $8.5 million in cost savings as of modified processes as of the fall of 2015.