The time required for bankers to decide whether to make a small business loan can be the difference between profit or loss.
A recent study by the Washington-based Business Banking Board found that traditional underwriting requires lenders to spend more than 12.5 hours processing a loan. With that long a decision-making process, experts say, poor margins are inevitable.
But manually processed credit decisions may soon be a thing of the past at most banks. Leading the charge to revolutionize decision-making on small business loans is Fair, Isaac & Co. In mid-March, the San Rafael, Calif.-based company will introduce its Small
Business Scoring Service, developed with Robert Morris Associates and 17 lenders. Broadening its reach, the company is also negotiating with the
Small Business Administration to develop a scorecard that would become the standard for quickly approving LowDoc loans nationwide.
The man leading the charge for Fair, Isaac is Latimer Asch, manager of the company's commercial products unit. He says the bottom line for banks that automate loan approvals will be minutes rather than hours for credit decisions. For decisions that can be fully automated, the whole process - including data entry, polling third-party repositories for information, analyzing the information, and calculating the credit score - takes about 15 or 20 minutes, Mr. Asch says. "Anytime you can take what is basically a 12.5-hour process and reduce it ... to a 15- or 20-minute process, the efficiency gains speak for themselves." In an interview after the Consumer Bankers Association's small business banking conference, Mr. Asch discussed the range of loans that credit scoring can be used for and how the scorecard will fare in recessions.
Q.: Some bankers are privately concerned that these scorecards have not had real-life testing under fire. Will banks be able to rely upon credit scoring during a recession?
ASCH: Traditionally, credit scoring models have done a very good job of adapting to recessions. What happens is the characteristics that make up the credit score, including a lot of information about the consumer credit report of the principals, business credit reports, tend to adapt their distributions within the characteristics to changes in the economy.
Q.: Give me an example.
ASCH: Take a situation where you have characteristics related to the trade payment performance of the business. Say you have a firm that, on average, pays its trade credit around 10 days beyond terms. If you hit a recessionary time, what we find is that the days-beyond-terms indicator tends to go up. That firm then will be penalized in that particular characteristic. Credit scores of other firms experiencing similar problems will also come down. You will realize subsequently, at a given strategic cutoff score, lower acceptance rates.
Q.: Is that proven or theory?
ASCH: This is based on about 30 years of empirical evidence derived from the consumer business that we've been involved in and about five years worth of empirical evidence from small business scorecards that have been in place in custom installations.
Q.: Where were these samples taken?
ASCH: We used a national sample with strong representation from thoughout the United States. As a matter of fact, we deliver not only overall score distributions but also regional score distributions based on the four large census region distributions.
Q.: Are there other ways scorecards can help lenders adjust for economic downturns?
ASCH: Not only does the score have a tendency to self-adjust, but the banks also control how they implement the scorecard and how they adjust their strategy. So if you do hit a time where you feel you have to pull back, you can increase the hurdle that companies have to get over in order to compensate for your perception of impending economic changes. And that's something that is much easier to do in a scoring environment than in a judgmental one.
Q.: Could you develop a scorecard purely upon historical information to see how it might react during a faltering economy?
ASCH: Anytime you develop a model, it is a balancing act. The only other way you could build a scorecard that included those recessionary times would be to go back further in time and construct a sample at a point in time further removed from 1995. The problem you would find is, yes, you would have downturns in the business cycle incorporated into the scorecard, but the financial indicators and the relative robustness of the data contained in the business bureaus is not particularly strong.
Q.: What is the best way to segment the small business market?
ASCH: One of the things we did when we developed a scorecard is, we went into detailed segmention analysis. We had sufficient data to develop multiple scorecards, so we looked at a statistically based analysis to determine what was the most effective way to split the population into multiple scorecards.
Some of the things we looked at were size of companies. Certainly, these are all small businesses, generally defined as having less than $5 million in sales. But we also looked to see the difference between companies with less than $1 million in sales and those with $1 million to $5 million in sales. We looked at different segmentations based on industry type.
Q.: And what did you find?
ASCH: As it turned out, the most predictable way to split the population was based on amount of credit requested. There were differences on industry types and by region. But there was a compelling technical argument to go for segmentation on the amount of credit requested.
Q.: Will this change over time?
ASCH: Every time we go through a system rebuild, we will perform similar analyses. As we get more and more data in and we get more and more systems in, it is possible that one of those other splits may come out as being preferred. But at the present time, we're showing that statistically the best way to split is based on amount of credit requested.
Q.: The scorecard Fair, Isaac developed with Robert Morris Associates allows banks to score credits as large as $250,000. What size credits can these scorecards legitimately handle?
ASCH: The development population included all those who requested credit of less than $250,000. A number of our users are not immediately going up to that limit. Some of these banks are saying that, during the rollout period, I'm going to limit it to credits of no more than $50,000 or $100,000. Some are going up to $250,000.
I can tell you I'm getting more and more pressure to extend it up to $500,000 or $1 million. Not only that but I'm getting a lot of pressure for extending this scorecard up into the middle market for use on a $20 million loan. We're looking at that.
Q.: Should banks avoid using the current scorecard for loans above $250,000?
ASCH: My recommendation would be to apply the scorecard to the same sort of a population that was used to develop it. If your population of applicants represents the same type of population as we included in development, the scorecard will work best.
If you take a population that included only requests up to $250,000 and try to apply that scorecard to $500,000 or $1 million deals, we just don't know how well it would work.
Q.: Is it possible to develop a scorecard on those $1 million credit applications?
ASCH: Yes, quite possibly we could. We're looking at the possibility of that, and certainly we've seen market demand out there.
But the big challenge in developing scorecards comes in developing the sample. Generally, we shoot for having 1,000 of each - we sample good loans, bad loans, and declines when we develop a scorecard. Almost always, the problem is finding a sufficient number of bads, especially when you're talking about loss ratios of 30 basis points. We can go down as low as 500 or 600 bads, but in fact, we've found even in the small business marketplace, there are at most 10 players out there that have enough samples to build a custom scorecard.
Q.: What kind of changes do you see in the future of credit scoring?
ASCH: One thing we've done is position this as a service. So we will continue to redevelop models. So it'll be evolutionary both in terms of the models as well as the software.
In addition to that, I see the next big opportunity coming in account management. That means being able to manage the existing accounts in your portfolio for renewal purposes, cross-sell purposes, to even handle those that go delinquent, collection prioritization.
We have sophisticated, state-of-the-art, industry-standard solutions on the account management end of the consumer marketplace, and we are looking to extend that technology to the commercial arena.