Ahead of Risk Evolve 2026, we spoke to Danny Dieleman about how markets evolve amid uncertainty, traditional valuation frameworks face mounting pressures from scarce or imperfect data. Danny Dieleman, Head of Wholesale Banking Capital Treasury at ING, explores how firms can adapt risk and valuation practices, leverage advanced analytics, and ensure defensible pricing.

Why are traditional valuation frameworks coming under pressure in today’s uncertain and data-constrained market environment?

There is probably more data available than ever in human history. Not only the amount of available data is incomprehensible, also the timeliness of the data is astonishing. We can obtain in many cases real-time data, or otherwise a few times a day. At the same time we are more and more relying on this data for key processes and hence we are more vulnerable to the availability of data and its quality. This certainly holds for financial models: they are used to mark-to-market transactions, for capital and risk management purposes, as well as other key processes in the bank. All models are in essence a simplification of reality and should therefore be treated with care. Certainly in areas where historical patterns are shifting, or where data by nature is scarce, skewed, or imperfect. 

Where are you seeing the most critical vulnerabilities in model assumptions as market structures continue to shift?

Despite the availability of data, there are still certain asset classes that remain relative illiquid and where data is scarce. But also liquid markets can turn illiquid in case of a market disruption and then instruments that can normally be priced by a simple look at a Bloomberg terminal can in the blink of a eye become difficult to price. The risk that you are running if data is difficult to come by is that your pricing can be way off, or alternatively, your price is correct, but you have a hard time explaining this to auditors, or your regulator. 

How should back-testing evolve to remain credible when historical data is less reliable?

Back testing will always remain an important tool to verify whether the model makes sense and whether the current calibration is still in line with the market. But, if observable prices are thin, or not well representative for the portfolio that needs to be valued, the information that is obtained from back testing will be less robust. As a consequence, it cannot be ruled out that there is more “model risk”, as either the underlying assumptions of the model, nor the calibration can be judged as adequate. This may have consequences for any model risk assessment that needs to be made as part of the model development and/or monitoring process.

What role can advanced analytics play in improving pricing transparency without adding unnecessary model complexity?

This is a tricky question, as advanced analytics has the risk of adding unexplained complexity to the model, and this is something that regulators, but also management is usually not keen on. There is, however, certainly a role for advanced analytics, as these tools can bring hidden dependancies to light, or they can help in areas where data is incomplete. This can be embedded in different ways in the model development process.

How are firms embedding adjustments to the modelled fair value into risk and valuation processes while maintaining consistency and defensibility?

This needs to be addressed in the governance of the valuation process. It is an important issue, as any adjustments, or overwrites to a model may come across as arbitrary, or cherry picking. So, a careful analysis, with input of all relevant stakeholders and a well documented process are some of the key elements to avoid any misconceptions. Also, too many overwrites of model outcomes may lead to the conclusion that the model is not performing well and a redevelopment needs to be performed. This may, or may not be desirable, but one needs to keep this in mind during the decision making process.