Q: What does Sageworks define as default? A: Sageworks only looked at monetary defaults, not covenant defaults. Default was defined as:
90 days past due, defined as 90 or more days past due on a loan payment
Nonaccrual, defined as loan being placed on nonaccrual status
Troubled debt restructure
Charge-off/write-down, defined as the reduction of the book value of an asset following the indication the value is diminished
Sageworks defined default as the earliest of the following events, so the dataset defined default as nonaccrual the majority of the time, providing a more conservative measure of default.
Q: How do I purchase more than one report or purchase a subscription? A: If you’d like to buy a bundled package or access an unlimited subscription, please contact Sageworks at email@example.com or call us at (866) 603-7029 ext. 691.
Q: On what type of borrowers or businesses should I run this report? A: You can run this on any business for which you have financial information. However, the dataset was built on small, privately held businesses; therefore, it is ideal for businesses under $10 million in annual sales. It can be used in the following ways: for corporate credit analysis, to factor into your risk rating policy as a quantitative metric, in loan review, in loan monitoring, or for loan decisioning and loan pricing. If you are unsure how a particular industry or concentration performs, this can provide objective data on the borrower and help with risk-based loan pricing.
Q: Is this appropriate for evaluating all types of loans? A: No, you should not run this on real estate development or construction deals because the performance of these loans depends contractually on cash flows from one development rather than on the financial performance of the business. Additionally, you should not use this for consumer loans because the Business Credit Report is only for businesses.
Q: Do I have to subscribe to other Sageworks’ solutions to use the Business Credit Report? A: No, you can use the Business Credit Report by Sageworks without subscribing to any other Sageworks products. If you are interested in subscribing to additional solutions by Sageworks, please contact us at firstname.lastname@example.org.
Q: I currently subscribe to Sageworks' products. Is the Business Credit Report included? A: The Business Credit Report will be part of the Sageworks suite; however, it is a separate product and can be added to your subscription for a fee. It can be used on a per report basis or purchased as a bundle or subscription. For more information on purchasing the Business Credit Report, please contact us at email@example.com.
Q: Is there any support or training provided on the product? A: If any users have questions on how to use the product, please contact Sageworks Support at (866) 603-7029 option 1 or firstname.lastname@example.org.
2: Using the Business Credit Report
Q: How long does it take to run a full report? A: Based on our research, the time to input the ten fields for a global analysis is on average three minutes. For a standard business-only analysis, the data entry would take even less time. This time does not include time to gather the data and calculate the factors needed. Focus groups with bankers and other users revealed that calculating EBITDA and Net Income took the longest off of tax returns.
Q: What if I don’t have all the pieces of information needed for the eight business inputs? Should I only fill in the ones I have? A: No, you should not use the Business Credit Report if you cannot fill in the eight inputs for the business. Leaving certain numbers as zeros will skew the results. Sageworks recommends getting those financials before running a Business Credit Report.
Q: Can I run businesses that only have interim financials through the Sageworks’ model? A: We do not recommend interim financials, but the model can handle annualized financials. The model was built on annual, year-end financials because year-end financials give the most accurate sense of the company, correcting for seasonality. Matching inputted data to that which the model was built on ensures the best results. You should annualize interim statements if that is all that is available.
Q: How does the model incorporate non-recurring or unrealized adjustments that need to be made? A: Our model does allow for adjustments based on your judgment of the borrowers’ specific financial situations. While we recommend that you only consider regular and dependable sources of income for covering debt service, if you believe that non-recurring or unrealized adjustments will impact the credit risk of a borrower in the coming 12 months, then you may modify the input fields to accommodate those adjustments.
Q: Would the model be able to handle a business that is rapidly expanding or just a stable business? A: The model can handle stable businesses, new businesses, or expansions. If you have financials for the expansion, you can enter that data into the input screens for the Business Credit Report if you believe those changes will better reflect the credit risk of the borrower in the coming 12 months. If it is a proposed expansion, feel free to tailor the numbers you input to reflect the expansion and how it would affect cash flow and debt in the coming 12 months. The probability of default will be based off the financials you enter for that borrower or business.
3: Conducting Global Analysis
Q: What should I do if I only have the business financial data? A: Sageworks provides two models: a business-only probability of default model and a global model. Both are predictive, consistently performing 79 percent better than publicly available models, such as the Z-score. Given the significantly predictive nature of Sageworks’ business model, you can use the business-only model if you cannot access the personal financial data or if you think the business-only will be a better measure for the borrower.
Q: I have a relationship with more than one business. How does global handle this? A: For business-only relationships, assuming that all the businesses guarantee the loan, you can run the business-only model and pool the inputs needed. To do this, add the specific inputs together for each business to arrive at a total number for each field. For example, to calculate sales, add sales together for each business in the relationship. You can perform calculator functions directly in the Sageworks input fields; there is no need to use an external calculator, spreadsheet, or paper.
Q: I have a relationship with a lot of owners and some of those owners are guaranteeing pro-rata. Does your model accommodate this, and how should I input this? A: Yes, the global model accommodates this case. Please weight the personal inputs that are under pro-rata guarantees with that owner’s ownership percentage.
Q: We often have businesses that have operating companies and real estate entities separated. How should we handle that in the model? A: Sageworks recommends combining the operating company and real estate financials if the real estate financials will affect the business’s ability to pay its debt in the coming 12 months.
4: The Dataset
Q: What are the ratios or factors Sageworks uses in the model? A: Sageworks uses the following ratios to calculate probability of default:
EBITDA: Assets (a measure of profitability)
Liabilities: Assets (a measure of capital structure)
Cash: Assets (a measure of liquidity)
Net Income: Sales (a measure of efficiency)
Debt Service Coverage Ratio (a measure of global cash flow)
Q: Of those five ratios, which are most important? A: Through testing techniques, Sageworks statisticians determined the most significant ratios for probability of default are liabilities to assets and cash to assets; however, all five ratios add to the predictive power of the model.
Q: What are the coefficients used in the model? A: The coefficients cannot be disclosed; however, the methodology whitepaper outlines the factors used, data cleaning process, and testing techniques in detail. Due to the proprietary nature of this model and commercial availability of it, we cannot disclose the coefficients.
Q: What does the methodology whitepaper mean by “predictive?” A: This is a model specification construct wherein the model was built on one year of data going in; thus, it can predict one year going out.
Q: Why have you only used financial data in building your model? A: First, financial data is a thread that connects everyone dealing with privately held businesses—it’s the truly objective standard. Thus, it forms the best source of data. Regardless of how long they have been in business or their size, privately held companies will have financial information if they are applying for a loan or line of credit. Second, financials are standardized i.e., “total assets” means the same thing whether you are receiving this number from a tax return, an audited financial statement, or from a borrower’s QuickBooks printout. Finally, financial data can capture the success, operational efficiency, debt service, capital structure, liquidity, size, profitability, among other characteristics, of a company and thus form a rich and flexible base for credit analysis.
Q: Did you initially look at other factors other than financial data? A: We did review other factors initially, including industry, state, zip code, and organization type. However, after robust testing, we found those not to be as predictive as the final financial ratios or factors in the final model.
Q: What about subjective factors? Numbers don’t always capture the ‘essence’ of a company. I personally know my borrowers—they’re good people but you’re saying that they might default in the next year. A: We don’t discount the value of qualitative factors e.g., relationships with your borrowers. We do believe that quantitative analytics provide a standardized, systematic report card for everyone. Root your qualitative evaluation on these quantitative analytics, not the other way around.
Q: Why do you only require one year’s worth of financials? I’m surprised that you didn’t ask for changes in the ratios from year to year. A: First, we found that building a powerfully predictive PD model did not require two years of financials. Second, since we are predicting one year of default, we only require one year of financials. Third, smaller businesses grow so quickly that we had extreme outliers due to quick changes, and the data did not suggest any significance in these percent change variables. Finally, we will continue testing percent change variables in future versions of the model, in case anything changes. Due to the adaptive nature of the model, we can add change variables if we find they become more predictive.
Q: Why isn’t industry included in the model? I would think that would impact a borrower’s PD. A: Industry was collected in the data gathering; however, we found it was not significantly powerful or predictive. The financial ratios took all of the explanatory power, and industry did not add to the predictive ability of the model. We believe industry is reflected in the financial data; therefore, industry is accounted for in the model through business financials that reflect how the business’ industry is performing.
Q: Why isn’t region or geography included in the model? I would think that would impact a borrower’s PD. A: Region, through state and zip code, was collected in the data gathering; however, we found it was not significantly powerful or predictive. The financial ratios took all of the explanatory power, and neither state nor zip code added to the predictive ability of the model. We believe location or geography is reflected in the financial data; therefore, location is accounted for in the model through business financials that reflect how a borrower’s region is performing.
Q: How are financials collected? A: Financial data were collected initially manually, by partnering with client banks of Sageworks and gathering their borrower data—both financials and loan information. Going forward, Sageworks is developing an automatic feed to continue collecting data from partnering banks to keep our dataset as up-to-date, relevant, and reflective of current economic conditions as possible.
Q: Will this dataset change substantially such that it is no longer representative? A: Our automated data collection process will ensure that we can update the model using financials from the latest, most representative dataset and the latest credit cycle events.
5: The Methodology
Q: Why doesn’t your product feature credit cycle adjustments? A: Sageworks’ model is built on the latest data through the credit cycle—the numbers already reflect the latest credit cycle conditions. As we update the model with the latest data, our model will automatically reflect the latest credit cycle conditions. We don’t have to create an “ad hoc” credit cycle adjustment like other products. Since we are sourcing data every day, our model incorporates that credit cycle information live.
Q: How is this model different from publicly available models? A: This model is different from other models due to our dataset. First, the nature of our data is different from other models and credit reports commercially available. It is financial statement data, not payment histories or community scores that can lag and be less predictive of default. Second, the data is based on small, privately held companies, a subset of the population that can be difficult to model and analyze. This dataset is not based on big, publicly traded companies and, therefore, is more reflective of the small, privately held companies users will be analyzing through the Business Credit Report. Finally, our dataset is more recent than other datasets that both publicly available models use and proprietary models use. Our dataset contains data from 2012 and is from post-recession information. Additionally, there are plans in place to continually gather data to ensure the model remains up-to-date, relevant, and reflective of current economic conditions.
Q: How is your product better than the Altman Z-Score or other publicly available models? A: The Altman Z-Score is an academic project built in the 1970’s on a small set of outdated data. The Z-Score is also “coarse” in that it splits companies into three “zones” i.e., the safe zone, the grey zone, and the bankrupt zone. This is because of its primitive statistical core: discriminant analysis. The Sageworks model with its Probit core is able to output more precise probabilities of default, up to four significant digits.
Additionally, Sageworks used ROC curves, standard in statistical analysis, to test the predictive power of the Sageworks model compared to other available models. For ROC curves, every single percent increase in the area under the curve (AUC) is highly valuable and difficult to attain. Thus, even a 2 percent performance increase for a model of this sort is significantly better. The Sageworks model outperforms the Beaver model by about 4 percent on the ROC curve and the Z-score by about 10 percent.
Q: Have you backtested the model? A: There are many ways to backtest a model. One formal method we chose was to perform out of sample testing, where we hold out a set of datapoints and test how the model performs on that held-out data. Another method we chose was to build separate models through time to simulate how the model would have performed when used across time. The results of both of these types of backtests, along with their area under the curve (AUC) metrics, are available in the model's methodolgy.
6: Industry Comparison
Q: How was data for the industry average sourced? A: We sourced a sample of privately held companies for each industry from our CPA and financial institution clients through our cooperative data sharing model. This set of privately held companies was run through the probability of default model, and an average for each industry was calculated.
Q: How often is the industry average probability of default updated? A: We update this industry average probability of default every 6 months.
Q: Do you provide average probabilities of default for 2 digit, 4 digit, and 6 digit industries (according to NAICS code)? A: Although we track industries down to the 6 digit NAICS code level, we currently only publish industry average PD statistics on the 2 digit level.