Extracting ROI From The Engaged Customer: A Portfolio Management Approach to CRM Magnify Analytic Solutions: Keith Shields, Chief Analytics Officer – Magnify, Chief Credit Officer – Loan Science Susan Arnot, Director, Decision Sciences Laura Benard, Director, Client Services Jen Boyer, Marketing Strategy Manager, Ford Customer Service Division
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Extracting ROI From The Engaged Customer:A Portfolio Management Approach to CRM
“Extracting ROI From The Engaged Customer: A Portfolio Management Approach to CRM”…
• What does the title mean?
• Portfolio Management, loosely, is the application of a set of collections and servicing techniques (typically analytically-driven) aimed at maximizing a loan portfolio’s cash flows.
• Why managing a loan portfolio, especially a student loan portfolio, is a CRM problem…
• Portfolio Management, loosely, is the application of a set of collections and servicing techniques (typically analytically-driven) aimed at maximizing a loan portfolio’s cash flows.
• Is it fair to define CRM as the application of marketing techniques (often analytically-driven) aimed at maximizing the repeat purchases of a set of customers?
• Loans can be thought of as bonds that throw off a stochastic series of cash flows.
• Customers can be thought of as bonds that throw off a stochastic series of cash flows, where that stochastic element is often estimated by statistical models that predict loyalty and retention.
• Uncollected debt can be placed in a variety of collections agencies based on which agency extracts the best cash flow.
• Wouldn’t a CMO be willing to “place” his/her customer portfolio with a new CRM entity (internal or external) if that entity could extract more repeat purchases from that portfolio?
• Champion / Challenger tests are very important, but serial analysis of Champion / Challenger tests can lead to the wrong view of the “big picture”.
• Isolated Champion-challenger tests measure enticement, not necessarily engagement.
• Enticement is measured cross-sectionally, whereas engagement is measured longitudinally.
• Measure success at a high-level (all the while inviting segmentation and drill-down). Make sure the measures are simple, with exact, and computed at over time:
• Example from auto parts and service: “20% of Jan13 servicers returned for service over the subsequent year. 15% of Feb13 servicers returned for service over the next year. 10% of Mar13 servicers returned for service over the subsequent year.” What’s the problem?
• A note: in general the most important metric associated with a loan portfolio is the sum of the cash it throws off over time: both as an absolute amount and relative to the original forecast.
• “Return Rate”, the % of Servicers Returning in the Next 12 Months is an exact measure, and it’s easy to track over time. Graph top right.
• The metric itself also invites segmentation; allowing insight into mix-shift and untreated customer populations. This sounds a lot like portfolio management.
• “Mix-shift” is a very important effect to understand. See the graph bottom right. How does this dynamic affect us when marketing to a portfolio of customers?
• From an analytics perspective, CRM and collections are essentially the same thing.
• In both cases you collect all the data you know about a customer at a point in time, predict likely behavior of the next 6-12 months, take action on that customer based on the prediction.
• Collections calls are the base treatment of portfolio management; private offers are the base treatment of CRM. Nuanced versions of those are left to champion / challenger testing.
• Infrastructure (DW, BRE) can and should be shared. www.zootweb.com
• => A full view of engagement: shared infrastructure allows real-time integration of CRM and collections.
• Per the Basel III Accord, a bank must know the probability of default (PD) for every loan on the portfolio. When the number of defaults exceeds forecast (PD*# active loans), then there can be a capital adequacy problem. This is a useful discipline…
• Shouldn’t CRM managers have a forecast of repeat sales / return visits? Seems like this comes directly from the loyalty & in-market models already in place.
• Lifetime Value Models are, in some sense, a statement about the worth of the company. When calibrated properly, they equal the net present value of the profit stream from a given customer.
• An eroding aggregated score from the Lifetime Value models can be symptom bad CRM.
• The “probability of return in the next 12 months” (PR12), is a model that can be applied to the servicer portfolio at the customer level.
• Aggregating the PR12 for each vehicle age segment allows us to predict the return rate for the segment.
• This puts us in a position to understand when return rates, and thus return visits, are higher or lower than we should expect…which in turn puts us in a better position to evaluate uncontrolled tests, like national rebate offers or ad campaigns. See graph right.
A national rebate offer in 1Q2012 creates 700 bps of unexpected response.
• When managing a loan we use the models and analytics to keep the loan in its most “valuable” state. Example: student loans, forbearance, reduced payment plans…
• Even small decisions, like the decision to place a collections call is, and should be, analytically-driven. Example: a pool of 1,000 loans are 15 days delinquent.• Contacting a 15-day delinquency reduces the probability of default from 5% to 4.8%. • We lose $5,000 for each default, so a contact is worth 0.2%*$5,000 = $10. • The contact rate is 5% => calling the 1,000 will generate 1,000*.05*$10 = $500 per day in
value.• We need three extra collectors to collect the 1,000 loans. Say collectors cost $6,000 per
month…roughly $200 per day. The additional three thus cost $600 per day.• $600 cost > $500 revenue => we do not call 15-day delinquent borrowers.
• How does this apply to customer engagement and CRM? See next slide.
• A CRM Manager should strive to keep customers in their most valuable state. Determining the most valuable state is often a matter of predictive modeling. For example (parts and service again):
• A customer requests $200 financial assistance with a repair that has occurred just outside of warranty.
• Loyalty and customer satisfaction models tell us that, given this particular customer’s demographics and past behavior, knocking $800 off the repair will increase his “satisfaction rating” from 3 to 5, which has the impact of increasing the likelihood of repurchase by 500 bps (5 percentage points).
• Putting the customer in a more valuable state (satisfaction=5) is worth 5% * $6,000 (the profit per vehicle sale) = $300.
• => The cost of putting the customer in a more valuable state ($200) is less than the benefit of having him there ($300) , so the assistance is approved.
• Understand that the job of CRM is to extract repeat sales and revenue from the portfolio of customers. The best way to do this is make sure that customers remain engaged over a long period of time.
• If a customer is a bond, then improving engagement, in effect, increases the life of the bond.
• CRM groups should measure themselves with this standard in mind.
• Keeping customers in their “most valuable state” is a matter of advanced analytics and strong marketing tactics…both of which are done with an eye towards engagement.
• The disciplines applied routinely to the management of loan portfolios are equally applied to CRM. Champion / Challenger tests are simply one tool in a larger toolbox.