Moneyball — how do I stretch my dollar?
Moneyball is probably an all time favorite movie for most Data Science lovers. And it accentuates the value that Data Science can bring to all walks of life — including sports… what’s not to love about it!
A line that stood out for me was the following nugget from Billy.
Billy Beane: There are rich teams and there are poor teams. Then, there’s 50 feet of crap. And then there’s us.
It symbolizes Oakland Athletics’ problem statement — in a league with no-salary caps — how do the Athletics fight against the rich dogs (Yankees, Dodgers with their annual $100MM+ salaries) and even the middle of the packers (the Twins, Nationals to name a few) and punch above its weight! What’s not to love about using data smartly! And as a fellow Baltimorean fan, I could really sympathize — being stuck in the same AL East league as the Yankees, Red Sox and even the Blue Jays).
Though not as dire as the Athletics (or the Orioles), FinTech/FinServices companies in the lending space look to get the most out of its data (just like the Athletics) and maximize the ROI as well.
High level look at various models used in FinTechs/FinServices with a deeper dive in subsequent posts.
- Acquisition:
- Credit Risk models are the bread and butter for lenders to keep losses under control. These determine who gets approved, their credit lines, and the product terms (APR & fees). Losses are the biggest drivers of expenses for most lenders, and lowering losses directly results in higher Net Income.
- Marketing Models are all about maximizing quality responders (not just basic responders who might get declined) with the limited Marketing investment. It entails distributing Marketing dollars across all channels — SEM, Social Media, Affiliates, Display, Letters, TV, Radio — to elevate the Brand and get quality customers to apply. Media mix modelling and Attribution models help to optimize the allocation of spend and inferencing. Response modelling helps to understand who are likely to respond.
- Fraud models are primarily used in 3 areas — to determine identity Fraud at Acquisition (normally by fraudsters but even in case of synthetic fraud), to determine transaction fraud (in case of, say, account takeovers by fraudsters), and payment fraud (by customers themselves to blow past their credit lines).
2. Customer Management
- As a customer continues to use the product, ongoing customer risk models are important to determine treatments such as Credit Line increases or term modifications (pricing, Credit Line decreases, account closures to name a few) for riskier customers.
- Balance Transfer models are important to generate more organic revenue from incentivizing high quality customers to revolve more balance and generate more interest revenue.
3. Customer Servicing
- Historically, Customer Servicing wasn’t seen as a key business area for company. However, with the focus on delighting a customer, providing them clean user journeys and great ability to self-service by moving them towards digital channels like the Web and Apps — CS has ramped up as a key area of focus especially with digitization of services.
- With the advent of optimizing call-center routing, smart IVRs and chatbots, customers can be serviced more efficiently — enabling agents to focus on the harder problems to solve for the customer and ideally with one call resolution.
- Staffing workforce as per the demand (intra-month, intra-week, intra-day) lowers agent idle time and also lowers abandoned customer calls. Automating multiple manual tasks or making them more efficient can also drive more efficiency, reduce errors and improve customer satisfaction.
4. Loss Mitigation & Collections
- Collections mirrors marketing ironically, given its myriad of channels — and the focus should always be on how to figure out who to target, with what channel and what offers can be made to the customer to result in a win-win (help the customer and lower your losses).
- Risk models, Payment propensity models and contact models are key to operate Collections efficiently as well as effectively in lowering losses.
The wide swath of opportunities in FinTech and Financial services makes Data Science — coupled with significant improvement in Tech in terms of data storage, newer models being deployed and higher processing speeds — an exciting tool in this area. To paraphrase Billy Beane — How can you not be romantic about Data Science in Fintech?
Source:
