Case Study: Optimising Credit and Collections Management at Edward Don

It is standard management practice to evaluate
performance in key business areas against industry norms. At Edward Don and
Company, we became quite concerned when our internal performance indicators
revealed that we were flagging 66.2% of our active accounts as carrying a high
credit risk. Subjectively, this number stood out as a very high scoring – and
therefore very suspect – result. In practice, the high level of customer credit
concerns were throwing up numerous road blocks in our core business, the
distribution of food service equipment and supplies in the US.  

Reacting
to what appeared to be a very poor customer base credit scenario
resulted directly in several impacts on our operations:  

• What in
  fact turned out to be an excessive number of orders were being placed on ‘credit
  hold’. This reaction had a negative effect of requiring our collectors to spend
  an excessive amount of time on reviewing and releasing orders, and therefore
  resulted in less time for value-added activities. They were primarily spending
  time on the phone contacting customers for payment. About 98% of held orders had
  to be released manually, resulting in execution delays and increased processing
  costs. 
• The knock-on problems included growing frustration among our
  sales force, and an increasing risk that our customers were not optimising their
  sales potential.  

Clearly, it would have been imprudent simply to
loosen our credit standards arbitrarily, which would run the risk of increasing
payment delays and counterparty failures because of inadequate credit analysis.
We needed to quickly find out which of our customers truly presented high levels
of credit risk, so that we could then more confidently accelerate order
processing for all the others. If we could achieve this, the vicious circle
would swiftly transform into a virtuous circle, with accelerated payment
performance, better satisfied customers, lower credit risk exposure, more
efficient operations and enhanced business development opportunities for our
sales force.  

The resolution of this issue was clearly a very urgent
corporate priority, as it was significantly impacting our core business
operations.   

The Base Situation

Edward Don and
Company runs a decentralised 41 person set of credit and collections teams,
distributed around our regional offices in Florida, New Jersey and Texas, and at
the corporate headquarters in Illinois, US. The team includes 35 credit
collectors/analysts, each of whom is responsible for about 1,200 accounts. 
There are about 35,000 active accounts, growing at the rate of 150-200 new
accounts per week. Our core business pattern consists of a high volume of
relatively low value transactions.  

Our original approach to credit
analysis involved the construction of generic scores, primarily based on
information supplied by specialist credit bureaux. These scores were used to
evaluate the creditworthiness of new accounts, so that credit lines and payment
terms could be assigned to each account. They were also used to monitor the
existing accounts, supplemented with the team’s analysis of published, financial
reports plus some internally originated account performance data. The team also
used the analysis to construct collection strategies where these were needed,
in reaction to seriously late invoice payment performances. 

Generally,
this describes a pretty standard methodology that is widely used in the credit
and collection business, but it did not seem to be generating sufficiently
accurate results for us. Therefore, the operation was becoming afflicted with
bottlenecks, and was increasingly stressed, as I have described above. 

The team felt that the poor predictive quality we were experiencing with
respect to accounts’ changing credit condition probably resulted primarily from
a lack of accurate and timely input data. The underlying strategy then in place
was, reasonably enough, quite conservative; but in practice, it was generally
tending to assign far too many customers to an inappropriately high risk status,
with the consequences I have outlined. One of the causes of this unsatisfactory
situation was the basing of the analysis of some accounts on the ‘ship-to’
location, rather than focusing on the actual legal entity that truly reflected
the risk. This approach will almost always lead to an underestimation of an
account’s credit status.  

Paradoxically, other flaws in the analytical
methodology were in some cases resulting in the underestimation of the true risk
that was being carried by some other accounts. 

So we were operating in
an unsatisfactory environment in terms of the timeliness and accuracy of our
credit management process, and were therefore experiencing growing problem
issues in collections, in risk exposure, and in related fields. We were also
becoming increasingly sure that we were by no means using our professional
credit and collections teams at anywhere near their full potential. Something
had to be done.   

Initial Solution Identification and
Validation

Our first analysis of the problem, and of potential
solutions, strongly suggested that we could obtain much more valid and valuable
results by using a statistical modelling process to support our credit and
collections operation.  

We decided to invite SunGard to perform an
initial analysis, and this they did in a complimentary validation exercise using
a proven statistical model solution called AvantGard Predictive Metrics.
Essentially, this model quantitatively predicts the chance that a given
customer’s current good payments performance will deteriorate at some point over
a time horizon of the next six months.  

In outline, the methodology for
the validation was based on the statistical behavioural analysis of our
historical accounts receivable (A/R) data. The end result showed the probability
of each account becoming delinquent, expressed for example as the amount of cash
at risk, which is a strikingly practical and vivid indicator of risk.
Accordingly, we could see ahead into areas of potential future risk, by being
able to evaluate the bottom line impact. 

As the predictions were derived
based on the analysis of real data that we had supplied, the team felt the
validation was based on an objective process that properly reflected our own
‘typical’ commercial operations.  

Among the innovations introduced into
the analysis were proven variables such as accounts payment (A/R) histories, the
impact of seasonal factors, the impact of changing economic conditions, and the
trend performance of actual orders. Our understanding was that the introduction
of such specific information would be effective and powerful factors in refining
the analytical process. 

The validation exercise operated on a
substantial volume of Edward Don and Company’s historical business data,
representing 18 months’ history of actual receivables collection. The analysis
reporting which we received was presented in clearly intelligible form, so that
our teams could see which accounts out of our entire set of client accounts were
the ones in reality, most likely to deteriorate in creditworthiness, and to
potentially become delinquent.  

The key to our internal validation of
SunGard’s model was the comparison of the forecasted results with the actual
delinquencies that Edward Don and Company in fact experienced. The results
showed a very high level of forecast accuracy and actual result correlation; so
we were quite effectively satisfied that the AvantGard Predictive Metrics model
had the proven potential to add real value to our credit and collection
operations.  

Solution deployment  

Interestingly
and most helpfully, the prospective solution validation exercise additionally
provided a reliable template for rolling out the live solution in our
country-wide operation, once we have received positive feedback from the initial
stages of the project.  

The implementation team that was eventually put
in place combined Edward Don and Company executives alongside SunGard’s
AvantGard Predictive Metrics product specialists. 

The key operating
result achieved upon launch was that the effective statistical analysis of each
month of our data, required just one business day’s effort from now on. After
this our teams were working with up-to-date, accurate and dependable data in
support of their core professional duties.  

This substantial change led
directly to several clearly quantifiable improvements being achieved in our
credit analysis and management, order management and collection processes. And
all of these advances had direct and positive impact on our core commercial
activities.   

Tangible Benefits Achieved  

I
have noted earlier that our prior credit and collection management processes had
flagged 66.2% of our customer accounts to be high risk; and this apparent
finding led directly to high volumes of orders being unnecessarily held up in
our order processing workflow.  

In contrast, our new tools enable 80% of
these orders to bypass the credit review system for immediate processing,
immediately enhancing our cash flow performance, and improving our customers’
opinion of our order execution proficiency.  Most dramatically, today only
14.1% of our entire A/R portfolio is now classified to be high risk. 

Another immediate benefit achieved was that our credit collectors / analysts
are now able to focus their professional attention where it is really needed,
namely on those accounts which have been properly identified to truly merit high
risk status. As our teams’ confidence in the quality of the new statistical
analysis grew with increasing practical experience, the credit collectors /
analysts found that they were able to dedicate much more of their time and
energy onto those account situations in which they could make a real, measurable
impact on improving our collection performance. 

It is also interesting
to relate that the process is now more highly automated because of our
confidence in the accuracy of the underlying analytical model. This means that
many customer orders now flow through the system without requiring any kind of
analyst intervention, and so are not impeded, as they used to be, by being held
for research and resolution. This operational improvement additionally reduces
our processing costs, enhances customer satisfaction levels, and liberates our
sales force to concentrate on business development.  

A further,
different type of benefit we have achieved is that, having identified the most
high risk customers, we can now simply and accurately cut the relevant credit
lines; and we can, conversely, increase lines to the more creditworthy customers
identified in our new analytical environment.   

In practice, we have
experienced a 25% increase in the volume of outgoing phone calls initiated by
our credit and collections teams, reflecting their new capabilities to be more
proactive in their account management duties, and in other priority operations.
 

We can now set account management strategies that are directly linked
to realistic risk assessment processes, both for the initial assignment of
appropriate credit lines to new customers, and also for prioritising collection
processes. The relevant information for each account is analysed and reported
using a simple six-point risk grading system, and we can initiate research and
report on any account, at any time. The reporting can also quantify the risk in
cash terms, to broaden the value of the analysis for treasuries and other
departmental units. 

Today’s still challenging economic and financial
conditions mean that we cannot afford to be in any way complacent in our outlook
toward credit and collection.  Every year, our team works with SunGard’s
AvantGard team to revalidate and recalibrate the statistical models that support
our operation, so that we can operate most effectively as external conditions
change.  

Conclusions