Principles That Guide Our Work at ValuRisk Partners

Now that I’ve launched my own firm, ValuRisk Partners, I’ve been reflecting on some of the lessons that have shaped how I approach model validation, and how those lessons will guide our work moving forward.

Here are a few principles I’ve learned over the years and aim to incorporate in every engagement:

🔍 Start with people:

Get to know the client, their culture, and especially the first-line modeling team. Understanding the expertise and intent behind a model is just as important as reviewing its technical merits.

📄 Do the homework:

Carefully review all model documentation - don’t rush it. Then go a step further: interview the stakeholders. It’s in those conversations that you uncover further rationale behind key modeling decisions, the context for assumptions, and the challenges they’re navigating.

⚖️ Strike the right balance between first and second line:

A healthy model risk management framework requires strong independence - but also open, respectful collaboration. Validations work best when the second line challenges modeling choices without undercutting the first line’s accountability. At ValuRisk Partners, we aim to reinforce that balance - not disrupt it.

📈 Performance monitoring isn’t optional:

Validations shouldn’t be one-and-done. Ongoing performance monitoring ensures models continue to behave as expected in changing conditions. It’s where a lot of risk - and opportunity - hides.

📊 Benchmarking adds depth:

Whether comparing a model to internal challengers, peer institutions, or external vendor tools, benchmarking can help validate reasonableness and flag blind spots. It’s not just about performance - it’s about perspective.

🧪 Back-testing matters:

Models that aren’t back-tested are models we can’t fully trust. It’s one of the most direct ways to assess predictive accuracy and understand limitations.

🔎 Models aren’t meant to 100% accurately predict the future:

As the British statistician, George Box, once said, "All models are wrong, but some are useful." Their real power lies in helping you understand your institution’s vulnerabilities - how your balance sheet might behave under stress, where assumptions could break down, and what decisions carry hidden risk. That’s the lens we bring to every validation: not just checking boxes, but helping clients better understand where they’re exposed.

✅ Focus on value, not just compliance:

A good validation supports safety and soundness. A great one also helps sharpen a modeler’s thinking, elevate governance, and ultimately drive better business decisions.