Credit underwriting
Credit underwriting is the process financial institutions use to evaluate the creditworthiness of prospective borrowers before extending credit. It helps lenders assess the likelihood of repayment and decide whether to approve an application, which supports responsible lending and risk management.
A new lender’s first credit policy draws on the regulator’s requirements, shared blacklists and credit bureaus, and the experience of lenders that have survived two or three credit cycles. Market limits offer a cautious starting point for new products. People help the team understand new customers and investigate fraud. The lender then writes down its policy, tests it against live results and tunes it as data comes in.
Start with the regulator’s rules
A new lender should start with the regulator’s rules for consumer lending and consumer protection. Write those rules first, reflecting what the regulator says about who may borrow.
Check the minimum age for borrowing and the minimum disposable income an applicant must have. Set out other requirements, including whether a job is required. In some countries, students may not borrow under the regulator’s requirements, so the rules should make clear whether they are eligible.
Use shared blacklists
Alongside the regulator’s requirements, connect to every shared blacklist available in your market, as well as credit bureaus.
Start with the lists you can access. You can later decide whether some are too restrictive or outdated. It is better to start with something than with nothing.
Copy established policy cautiously
Beyond regulator rules and shared lists, many new lenders look to the policies of established lenders. A new lender may hire a credit analyst from an established company and copy its rules, even though the new team may not know why they exist. That can help prevent the lender from blowing up on day one, but it can also leave approval rates too low. A lender may choose to be risk-averse at the start and become more lenient later.
The bigger mistake is assuming your policy will be better because you are smarter than lenders that have been in the market for 10 to 20 years. They have been through at least two or three credit cycles and know how business and credit cycles work. If they have survived, something in their credit policy makes sense. Thinking you are smarter than them is a stupid assumption.
Set first limits against the market
A loan is sized from the applicant’s disposable income, the lender’s financing capabilities and its funding term. The monthly payment has to fit what the applicant can pay, and the loan term has to fit within the period for which the lender has funding. If the money is available for three years and the product runs for two, the terms fit.
A first credit card limit starts from a different comparison. When entering a new market, look at what established lenders offer. Do not set a limit above what most other lenders have given someone.
Credit history helps show what limits a person already has with other lenders and how much of each limit they use. Credit records show each account’s balance, status and payment history, and usually its limit. Some issuers leave the limit out, so the picture may be incomplete. A person may already have cards with several companies, and their existing credit lines show both the limits lenders have offered and whether those limits are being used.
Some lenders logged in programmatically to the applicant’s account at another lender, then matched the limit or added 10%, for example.
Under PSD2, the European Banking Authority’s final standards banned screen scraping, where software logs in to a customer’s online account as the customer. Since 14 September 2019, banks in the EU have had to provide licensed account information providers with a dedicated interface, used with the customer’s consent.
Keep people close to new customers
Market comparisons do not tell a new lender everything it needs to learn about its customers. Welcome calls and sensitivity tests involving new customers should never be automated. When entering a new market, have a few credit people make welcome calls to new customers. People should also handle some sensitivity tests involving new customers. The calls help the credit team work out who the customers are.
Fraud investigation needs people who know how to investigate cases. They need to work from the data and find out what is really happening. Underestimating that work becomes a problem when a fraud bubble hits or fraud hits the lender.
Write down the policy before automating
Moving a credit policy from a spreadsheet into a system that decides live applications can take hours or days, depending on the decision engine. With an old, complex credit decision engine, it may take a few months.
The harder case is manual underwriting with no formalized credit policy. The written policy and the way each loan is actually approved can differ sharply. A lender may discover that loans are being approved very differently from what the policy suggests. That gap makes automation difficult. If the policy has never been formalized, the work can become a mess.
Make stability the test
Live applications then show whether the written policy produces steady results. In the first three months, you book the first loans and get the first payment default. If that first payment default is somehow stable, keep checking how the policy works over time.
With a statistically significant number of loan applications, the approval rate and bad rate should stay stable from day to day. If the approval rate is 60 or 70 percent, it should remain around that level each day. The bad rate should also stay consistent. Check it daily or weekly, depending on how many loans you book. With enough applications, it should not be 3 one day and 10 the next.
As data comes in, tune the policy by industry segment, the lender’s capabilities and how collections are performing.
Questions we get asked
What should set a new lender’s first credit-card limit?
Anchor the first limit to what established lenders have already offered, not to a higher limit you would like to offer. Credit history can show existing limits and how much borrowers use, but some issuers do not report limits. Where the record is incomplete, start cautiously and revisit the limit once your own repayment results are available.
When should a new lender start changing a policy borrowed from established lenders?
Use the first three months to book initial loans and observe the first payment defaults, then wait for approval and bad rates to hold steady across enough applications to judge them. Check daily or weekly, depending on booking volume. If the rates jump around or the numbers are too small to read, keep gathering results rather than changing a cutoff on the strength of one day.
What needs to happen before we automate manual underwriting?
Write down how credit decisions are actually made, then compare that policy with the rules people say they follow. If the two differ, resolve the gaps before building the decision logic, or the system may encode an inaccurate version of the policy. This work can take longer than moving a documented policy into a decision engine.
The credit policy management page shows how credit policy changes become revisions that teams can compare, test again, and assess for their impact on past decisions.
