Automated KYC Identity Checks in a Loan Application: Flow Design and Cost Model

A lending company wants every loan applicant's identity confirmed automatically during the application, rather than by staff checking documents by hand. As the product manager, design this identity-verification feature, and then say how much it would cost. ### Clarifying Questions - Is the applicant a consumer borrowing personally, or an owner applying on behalf of a business, in which case the business and its owners may also need checking? - Which countries does the lender operate in, and which identity documents and data sources are available there? - Is the main driver regulatory compliance, lower fraud losses or a smaller manual-review workload, and roughly how many applications arrive each month? - When the interviewer asks about cost, do they mean the cost to build the feature, the cost to run it for each application, or both? ### Part 1 — Design the automated identity check Describe what the feature checks, where it sits in the application flow, and what happens to applicants who cannot be verified automatically. ```hint Not every applicant needs the same check Some signals cost the applicant nothing and others add effort. When is the extra effort worth asking for? ``` #### What This Part Should Cover - The fraud and compliance risks the check must catch, and the cost of wrongly blocking a genuine applicant. - The verification signals and the order in which they are applied. - The decision outcomes, including a manual-review fallback and a path for people the automated checks cannot handle. - Success metrics, with guardrails on applicant drop-off and fairness. ### Part 2 — Estimate what the feature costs Put a number on the feature. Show how the cost is built up and how it compares with what the lender gains. ```hint Once, then per application Separate the one-off cost of shipping the feature from the costs that grow with every application, and remember that some applications will still reach a person. ``` #### What This Part Should Cover - A cost equation with every input labeled as an assumption. - One-time build or integration cost versus recurring per-application cost. - A build-versus-buy decision for the hardest components. - A comparison against the benefit, such as fraud losses avoided or review hours saved. ### What a Strong Answer Covers - A design anchored in the lender's goal: stop impostors and fabricated identities while keeping genuine applicants moving. - Consistency between the two parts: each verification step from Part 1 appears as a cost line in Part 2. - Handling of personal and biometric data: consent, storage, retention and access. - Awareness that error costs (fraud let through, good customers turned away) can matter more than vendor fees. ### Follow-up Questions - After launch, pass rates at the selfie step drop sharply for applicants from one region. How would you investigate? - Fraudsters begin passing the document check with high-quality fake IDs. What would you change first? - Finance asks you to halve the cost per application. Where would you look?