From Model Development to Model Confidence: What Leading Banks Do Differently
Across the earlier articles in this series, I’ve been trying to unpack where modelling challenges are often misdiagnosed. I began by questioning the assumption that the most persistent difficulties sit in the algorithms themselves. From my experience, that is rarely the case. I then looked at governance and how I often see it framed as a barrier rather than the structure it is intended to provide. From there I explored the model lifecycle, and how fragmentation across that lifecycle introduces friction that only becomes visible later on. More recently I focused on explainability and fairness, not as compliance requirements, but as design choices. In practice both play a significant role in determining whether understanding and intent are preserved as models move through an organisation. When I step back and connect these threads, they lead naturally to the same place. Not simply better models or more efficient processes, but something more fundamental.
Confidence.
It is a concept that rarely features explicitly in technical discussions. In practice I hear far more about performance, accuracy, stability and robustness. Models are assessed through measures of predictive power and statistical significance, and rightly so. Yet when I observe how decisions are actually made within banks, confidence plays a much larger role than is often acknowledged. I’ve seen models that perform well on paper and still struggle to progress if those reviewing it are unclear about how it was built, cannot easily trace its logic or feel uncertain about its long‑term sustainability. When that happens, momentum slows, questions multiply and reviews become more prolonged and cautious.
By contrast, when a model is clearly understood, well documented and supported by consistent evidence, the tone of the conversation changes. Discussions are more focused, concerns are easier to resolve and decisions tend to be reached more quickly. In both cases, the difference is rarely the model itself but the level of confidence that surrounds it. This dynamic becomes particularly evident as models move through an organisation.
During development, confidence is often high. The team building the model understands the data, the assumptions and the decisions made along the way. Everything feels coherent and internally consistent. As the model moves into validation and review, that context naturally begins to thin out. New stakeholders become involved, including risk teams, validators and auditors, each approaching the model from a different perspective. Their focus extends beyond what the model does to how it does it and whether it can genuinely be relied upon. When clear answers to those questions are not readily available, confidence starts to erode, and once confidence dips, progress almost inevitably slows with it. What I find interesting is that leading banks appear to approach this challenge differently. Rather than concentrating solely on how models are constructed, they pay close attention to how confidence is established and maintained throughout the lifecycle. Development is treated not as a discrete technical task, but as the beginning of a longer journey.
Decisions are not only made but captured in a way that allows them to be understood later. Assumptions are clearly articulated. Data transformations and feature choices are traceable and supported by context rather than memory. This approach extends well beyond the development phase. Documentation, review and approval are treated as integral parts of the same process, with evidence built up progressively rather than assembled under pressure at the end. As a result, validation becomes less about uncovering gaps and more about confirming what is already visible.
Over time, this creates a different experience for everyone involved. Instead of having to reconstruct the model during review, stakeholders are able to follow a narrative that has been developing from the outset, making challenge more constructive and resolution more straightforward.
I also see a clear shift in how these organisations think about governance. Rather than being applied retrospectively, governance supports the entire journey. It provides structure, consistency and a shared set of expectations, making it easier for different teams to engage with the model with confidence rather than caution. This is especially important in regulated environments where models must withstand scrutiny from multiple angles. Confidence is not only an internal requirement but something that must be demonstrated to auditors and regulators as well.
When these elements come together the impact is tangible. Models move through validation with fewer surprises. Audit discussions are more focused. Decisions are made with greater clarity. Most importantly, models are far more likely to reach production where their value can actually be realised.
From the outside this can look like efficiency. In practice it reflects something more fundamental: an environment in which confidence is embedded into the process itself.
This brings me back to the broader theme that runs through this series. The industry often focuses on how to build better models. That question matters but it is only part of the story. In reality, success is rarely determined by the model alone. It is determined by whether an organisation can develop it, understand it, govern it and ultimately trust it.
The shift required is subtle but important. Away from models in isolation and towards the systems that support them. Away from performance metrics alone and towards confidence. Away from isolated development and towards a connected, well governed lifecycle. Seen in this light the earlier ideas begin to converge. The model matters. Governance matters. The lifecycle matters. What ultimately ties them together and allows models not just to be built but to be used is confidence. Because once confidence is in place, many of the challenges organisations struggle with begin to fall away. And that is what allows models to deliver value.
Written by Jalal Khoylou, co‑founder of Paragon Business Solutions, working across the credit industry.