The Cost of Getting Underwriting Slightly Wrong at Scale

Underwriting is built around assumptions. 

What does a borrower actually earn? How stable is that income? How much existing debt can they service? What loan amount can they reasonably repay? 

No underwriting framework will get every assessment exactly right. The real concern is when a small gap in those assumptions consistently points in the wrong direction. 

At scale, even a modest underwriting bias can materially change portfolio outcomes. 

It Starts With a Small Difference 

Suppose a lender consistently overestimates income stability for a particular borrower segment. 

The difference may be small enough that individual cases do not immediately stand out. But if the same assumption influences thousands of credit decisions, the resulting portfolio can look very different from what the lender originally expected. 

The same applies to loan sizing, debt obligations, bureau interpretation or other inputs used in the credit decision. 

The issue is not one bad loan. 

It is a systematic tendency to make slightly optimistic decisions. 

Where It Shows Up 

The impact may not appear in the underwriting itself. 

It can emerge later through higher than expected early delinquencies, lower repayment performance or credit costs that consistently exceed the assumptions used in the original portfolio projections. 

This is why simply tracking overall portfolio delinquency may not be enough. 

A lender may need to look at performance by customer segment, geography, sourcing channel, ticket size and other characteristics that were relevant to the original credit decision. 

The objective is to identify whether the problem is random or whether a particular underwriting assumption is repeatedly producing the wrong outcome. 

Turning Portfolio Performance Into a Credit Signal 

This is where the relationship between underwriting and portfolio monitoring becomes important. 

Once repayment patterns begin to emerge, lenders have an opportunity to test whether their original assumptions were actually predictive. 

If borrowers who were assessed similarly are consistently behaving differently from what the model or policy expected, that is useful information. 

It may point to a variable that needs greater weight, a threshold that needs recalibration or a segment that requires a different approach. It may also reveal that the original assumption was based on incomplete or outdated information, or that borrower behaviour has changed since the policy was designed. The next step is to determine whether the issue is isolated or systematic, understand what is driving the difference and update the relevant model, policy or monitoring process accordingly. 

Good underwriting therefore does not end when the loan is sanctioned. 

The portfolio itself becomes a source of evidence for improving the next credit decision. 

Why Scale Changes the Equation 

A small underwriting weakness can remain difficult to notice in a small portfolio. 

In a large book, it becomes harder to ignore. 

The difference between expected and actual performance, even if small on an individual account, can translate into significant credit costs when repeated across thousands of accounts. 

This makes underwriting accuracy particularly important during periods of rapid growth. A policy that appears to work at one scale may expose its weaknesses as the portfolio expands or its borrower mix changes. 

The goal is not to create an underwriting process that never gets a decision wrong. 

It is to create one that recognises its own blind spots and gets better as portfolio evidence accumulates. 

At scale, underwriting risk is often less about one bad decision and more about a small assumption that keeps producing the wrong outcomes.