Segment-level Portfolio Behavior and the Science of Forecasting: Current Practice and Emerging Promise
Signals, Spring 2003
Ultimately, any forecast benefits greatly
from the identification of groups of
customers or accounts exhibiting
different timing of lifecycle events and
divergent responses to environmental
conditions. Consumer portfolio analysis occurs largely along familiar lines:
- Total portfolio results (the p&l)
- Segment analysis for channels or sub-products
- Vintage analysis, a form of segmentation
- Account-level or very small cohort analysis.
While forecasts are often derived and reported at such levels,
we contend that projections become much more accurate when the
underlying portfolio segmentation is tailored to the forecasting
problem. A more rigorously defined segment analysis delivers a
better understanding of the portfolio, its likely direction and
its potential reaction to shocks. Such understanding assists senior
managers to make decisions anticipating expected future returns,
economic capital requirements, response to competitive threats
and other strategic issues.
Despite the promise, few organizations pursue segmentation as
a means to achieve better visibility and forecasting accuracy.
Most often the choice of portfolio segmentation for forecasting
is driven by preferences for familiarity or a corporate-reporting
process that emphasizes product or channel. But with corporate
commitment to new analytic technology, segmentation optimized
for forecasting can deliver on its promises, particularly in the
realm of risk-adjusted profit, portfolio predictability, early
warning, and the pursuit of beneficial portfolio effects that
reduce volatility and increase value.
Outlined below are current industry approaches to segmentation
from the most basic practice to the emerging frontiers, examining
the promise and potential pitfalls that can occur when segmentation
meets the forecasting problem.
Segmented analysis allows managers to view the portfolio at an
acceptable and sometimes intuitive level of aggregation. Common
practice finds portfolios sliced along certain management realities
of running the business: existing or historical partner channels,
large portfolio acquisitions, and product or program features.
Many segmentation schemes in use today are based largely on origination
channel (e.g., broker vs. branch-source), or reflect product types.
Others reflect geographic location of customers or branch locations,
which can be tapped to reflect variation in local economic conditions.
While these segments can often deliver fine forecasts for some
portfolios, they can also result in over-segmentation if your
forecasting goal is to understand issues with a low event rate
like charge-off or bankruptcy.
Analytically-based segmentation schemes typically support customer-level
decisions and associated management information structures in
consumer credit marketing and risk. On the risk side, segmentation
is frequently influenced or created directly from score bands
or some other measure of credit quality such as loan-to-value
in real estate loan portfolios. These actions can be useful for
credit analysis but often do not address issues of portfolio value
or the timing of losses (sidebar, right). Similarly, marketing
segmentation methods such as lifestage, demographics, and psychographics
look to optimize contact, response and utilization rates but are
not often employed beyond the point of origination.
In response to the credit risk and marketing-centric views illustrated
above, institutions have developed profit and behavioral-based
segmentations to try to identify, originate and develop high-value
customer relationships. This is a sophisticated approach that
requires excellent customer understanding and good information
on internal costs and in some cases a total view of the customer
relationship. While a value-based segmentation can provide a better
answer, there can be issues with their robust application. One
still runs the risk that some selection bias has occurred in developing
the segmentation (and therefore the forecast) based on economic
influences or reliable readings on the quality of originations.
For each of the segmentation strategies outlined above, a good
understanding of the moving parts in a portfolio—parts like
variation in origination quality, stability of the customer lifecycle,
and the positive or negative bias introduced by economic conditions—brings
clarity to historical portfolio performance. Each of these moving
parts has unique patterns or signals as they occur in portfolio
histories and as they project into the future. For accurate portfolio
forecasting, the moving parts are not an annoyance but should
be viewed as authentic contributors to a comprehensive picture
of the portfolio going forward.
Advanced analytical and optimization techniques found in Strategic
Analytics' Dual-time Dynamics (DtD) technology can help identify
and quantify the moving parts of a sound portfolio forecast. Coupled
with the adoption of a segmentation strategy specific to the forecasting
issue at hand, the combination delivers unrivaled portfolio visibility.
For example, segments of customers sensitive to the influence
of economic changes, and customers not as sensitive can be identified,
allowing the forecast to reflect the very distinct impact economic
changes will have on future portfolio performance. Ultimately,
any forecast benefits greatly from the identification of groups
of customers or accounts exhibiting different timing of life-cycle
events and divergent responses to environmental conditions. This
dynamic clarity is at the heart of DtD's contribution to the forecasting
problem.
So what is the bottom line for having an "optimal forecasting
segmentation" and better portfolio predictability as a result?
Portfolio intelligence, that is, knowing how the portfolio will
perform under different conditions and scenarios, gives the institution
a better handle on capacity planning in collections, a clear picture
of economic capital requirements and resulting investment decisions,
and the ability to better forecast account groups used in securitization
branches. This next frontier in portfolio analysis ties sophisticated
measures of segment value directly to an ability to predict behavior
accurately over the customer lifecycle and across economic cycles.
Portfolio managers will then know with more certainty which levers
to pull and which knobs to turn in order to move the portfolio
in a given direction: toward higher value.
|
Top of page