Modernize your D&A strategies: Temporal demand

The demand for timely data and faster processing has grown exponentially for financial institutions in recent years; however, many organizations struggle to select, establish and operate the appropriate processes and technology to ingest data in real-time and run analytics on-demand. This temporal demand for data ingestion and processing can empower firms with faster extraction of data, reducing manual intervention in reporting, faster deployment of models, better predictability of future outcomes, thereby improving better decision-making process and improving bottom line, if done correctly. When done incorrectly – or inefficiently – this struggle impacts all components of the data and analytics value chain and across all functions of the institution. In this post, KPMG discusses how firms can leverage the right processes and technologies to yield an effective temporal demand that sustains the data and analytics value chain.

Before understanding how to increase value, it is important to recognize the continuous dependency between data and analytics on temporal demand. If a process, for example a company’s financial planning and analysis (FP&A) forecast, aims to provide weekly model results to track progress, then it is essential that the integrated models that make up the FP&A take no longer than a week to run – ideally even less. But that just covers the temporal demand on modeling and reporting. To provide real value, the information going into those models must also be updated at least weekly. Input data that relies on long-run batching might be updated only once a month (or quarter), rendering the FP&A results stale. And these unbalanced dependencies can further spiral. Some data may be reliant on upstream models. And if those models take a few days to complete, then the data must wait, and the FP&A must wait even longer.

The above example takes a simple, yet ubiquitous process across companies, and shows the precision and balance needed to address temporal demand. When we think of highly regulated industries, such as banking, lower temporal demand can be risky as well as costly. For instance, pricing and interest/market rate models can expose banks throughout the income statement and balance sheet.

• Throughout the banking industry, pricing models have typically relied on market information for interest rates and static credit rating scales. But there are risks and costs to these strategies. Following market trends in rates can cause a bank to lose out in rate declines. And static credit assessments may expose a bank to unbalanced risk concentrations or miss out on underserved borrowers. Advancements in interest rate forecasts and credit risk analysis have allowed banks to turn these missed opportunities into revenue gains, providing actionable information sooner and clearer.
• In banks, model monitoring is typically performed quarterly or annually, which creates risks under highly stressed and volatile conditions. This risk is further increased for models that need to be run more frequently than they are monitored, such as liquidity risk. From tracking model degradation to making routine updates, model changes during these conditions can increase the risk around information quality and decision-making, information and decisions that can make or break a bank. Achieving real-time continuous improvement and deployment (CI/CD) through AI/ML and platform optimization is critical to address increased temporal demand under stress and volatility.

As banks begin adopting more powerful and integrated platforms, particularly in the cloud, KPMG sees a significant shift in temporal demand. And this shift directly drives business value, as banks are able to become better informed, lower their costs and establish strategic visions.