Background

Scorecards are a staple in the world of credit risk assessments to help make consistent decisions with less manual effort. Yet they have certain disadvantages that are often overlooked and need to be carefully considered when choosing a methodology for credit risk assessments in particular contexts.

In the first part of this series, we examined the pros and cons of Business Information Reports. Like BIRs and many other financial tools, scorecards come with their own set of benefits and limitations that warrant a closer look. In this article, we specifically explore trade credit scorecards, shedding light on their benefits, drawbacks, and which businesses should be using them.

What is a Scorecard?

A scorecard is a standard set of rules that balance certain positive and negative elements within a credit application, generally producing a single score or risk category for easier decision making. These rules are often implemented in a simple spreadsheet or in software. The rules and weightings are often adjusted to suit the lender/supplier’s risk appetite with the outcome of the scorecard automatically qualifying or disqualifying (‘knock out’) an applicant based on the balanced scorecard.

Businesses create and manage their own scorecards, but the following rules are common:

Each rule is weighted and added together to produce a combined overall risk score.

That sounds simple enough. But why doesn’t it work? The shortcomings are described below.

The Five Shortcomings of Scorecards

The Specificity Problem

Scorecards are “one-size-fits-all”

All customers are processed according to the same rules. And yet there are very big differences in how each rule might apply to them. For instance, a profit margin rule will favour certain industries that have higher margins but this might not align with the business strategy. Some sectors are known to have thin margins but operate at very high volumes, such as logistics. Comparing a logistics customer to a manufacturing customer based on profit margins doesn’t accurately reflect the relative risk. And it probably doesn’t reflect the business strategy to target customers in a particular sector.

To compensate for this, the scorecard would have to be setup separately for every industry. But the same segmentation problem exists for other areas too. For example, a new company has different dynamics than an established player. If the scorecard favours established businesses only, then you will do very little business with new market entrants and over time your market share will deteriorate. So you may need different scorecard for new companies. That means two scorecards (new and established companies) for every sector. But what about international companies? What about segmenting according to payment behaviour. The list goes on.

So, there will be an almost infinite number of ways to segment customers and therefore the need for an endless array of scorecards to accurately assess according to your risk appetite per segment.

Trade Shield

Trade Shield takes a statistical modelling approach – it does not use scorecards. Effectively this means that the modelling process takes all the potential customer profiles into account (hundreds of thousands of profiles) and develops the precise weightings for every customer profile type. For each of these profiles it determines the probability of default and late payment and allows customers to set the desired risk appetite for each using a simplified framework, called the Risk Adjustment Framework. The personalised handling of clients’ unique contexts builds customer relationships.

Trade Shield, to put it simply, enables the customization of hundreds of thousands of scorecards in one simple process. And it has further advantages over scorecards, discussed below.

The Sampling Problem

Scorecards address limited circumstances

Scorecards are built using rules-of-thumb and years of risk assessment experience. They represent the average rules that assess common scenarios (technical term would be a small common set of samples), but not other scenarios. Weightings are usually derived by only checking these common scenarios leading businesses to automate a process that only works for a limited set of scenarios.

But how often are risks arising from uncommon scenarios? Even if 75%of the risk is picked up in these common scenarios, it leaves far too many risks undetected.

Trade Shield

Trade Shield monitors the entire ‘data universe’ and the modelling process assesses risk across all features that are determined to be predictive, developing the ability to identify risky scenarios that are uncommon to humans but predicted by the data.

The Rules Source Problem

Scorecards are limited by human intuition

Human intuition is great in certain circumstances but generally unreliable and biased when it comes to credit scoring. In developing scorecards, humans propose the rules and the weightings, but they miss data patterns and signals and remain biased by what they already know to be “true”.

For example, we have proven that trade references are not good predictors of future default. And yet they are commonly considered in scorecards because they are simple to understand and have been used traditionally! Unfortunately, a good trade reference can be given right before a company enters bankruptcy because the supplier was essential. And a bad trade reference can be given when a reasonable dispute around quality results in withheld payments. There is simply too little signal among all the noise for it to be truly predictive.

So, scorecards include rules that are not predictive because of human bias. And they exclude rules where humans cannot analyze the data (because it takes massive computing power to extract these rules into a model).

Trade Shield

Instead of relying on humans to choose the rules and weightings, Trade Shield relies on powerful statistical techniques to remove bias and extract a predictive model from the data itself. With a team of data scientists unleashing these techniques on a massive dataset with nearly unlimited computing power, the models produced are astonishingly powerful, far outstripping the combined capabilities of every credit executive and credit analyst on the globe.

The Adaptation Problem

Scorecards are hard to update.

They do not automatically update their rules and weightings when circumstances change – which is often. Internal changes such as shifts in strategy and product mix and financial capability all affect credit risk tolerance and credit strategy. And external changes in the economy also drive the demand to update your credit risk appetite in particular ways.

With scorecards, all changes need to be made an ‘expert’ in response to a change in circumstances. But often the process to approve these changes takes so long, their effectiveness is severely diluted, the crisis has passed and a new one is already on the horizon.

Leading businesses aim for better decision-making, which is only possible when their decision-making framework adapts correctly and fast enough to be useful.

Trade Shield

Trade Shield’s predictive models are retrained continuously based on all new data received, to incorporate the most up-to-date modelling strategies. The approval process is simple – does the model improvement have greater accuracy in prediction? This enables Trade Shield’s recommendations to be based on a near-real-time framework, faster than any organisation could achieve with a scorecard-based-framework.

The Forecasting Problem

Scorecards don’t provide a view of the future

A scorecard is based on rules of thumb that are the average across customers. It looks at the past, and takes an average and does not predict the future. Its like saying, the last 5 days it was sunny, so tomorrow is likely to be sunny. In credit terms, scorecards are poor at forecasting expected credit losses because this needs a model that can predict future outcomes depending on certain scenarios.

Trade Shield

Trade Shield’s default and payment risk models accurately calculate expected credit losses for the next 12 months. This enables the forecasting of expected credit losses across a portfolio of customers. Such forecasting enables valuable decisions to be taken, as you can assess the impact of short term and long term credit extension on profit margins and cashflow, instead of relying on rules-of-thumb averages of the past.

What should your business be using?

Scorecards are suitable for your business if:

Trade Shield is suitable for your business if: