Fast and slow data: How to enable fast, interactive customer journeys based on slow mathematical models
  • February
  • 2022

Fast and slow data: How to enable fast, interactive customer journeys based on slow mathematical models

When it comes to digital journeys, one characteristic defines quality beyond industrial specifics: speed. While rule-based apps or websites are relatively easy to keep lean and quick, the financial industry may be the area where the speed of underlying calculations could be an issue. Unlike e-commerce or media, the digital and physical solutions provided by the financial sector are riddled with computationally heavy models trying to grasp the uncertainty of real-world economies. The more granular and elaborate the underlying model is, the more realistic and accurate its results are. Does it mean that the financial institutions will have to compromise on quality to deliver fast solutions?

Today we have spoken to Erik Brodin, an ex-McKinsey quant expert at Kidbrooke, who doesn’t believe a compromise is necessary.

  • Thank you for agreeing to talk to us today, Erik! When it comes to financial services, how important is it to optimize the speed of the customer journeys? 

Computational speed is one of the core considerations when creating interactive financial journeys, either fully digitalized or assisted by a physical financial adviser. If a part of the journey takes too long to compute, there is a significant risk that a customer becomes frustrated or loses interest in completing it. That’s understandable - no one likes to wait for a page to load!

Wealth management, including investments and pensions, is an area that requires information gathering, consideration of customers’ risk and economic profiling and using mathematical models. There are two ways to achieve seamless customer journeys in the financial services sector: opting for fast and simplistic models or cleverly using more complex models.

  • The simplistic models seem to be a more popular choice, aren’t they?  

A simplistic model can be computed quickly but will most likely have a considerable model risk due to not being realistic, i.e giving bad advice. Further, if one combines many separate simplistic models which power different aspects of financial advice, one can end up with results that contradict each other. 

At Kidbrooke, we have chosen not to take this route as we believe that a solution consisting of multiple simplistic models will not help the consumers be comfortable with the future of their financial situation.

Whether you are building an automated solution providing financial advice or looking for better tools as an adviser, ensuring that the model fueling all use cases is as realistic as possible determines the quality of the results. High-quality models are instrumental to providing consistent and sensible advice. However,  a well-crafted customer journey requires careful engineering.

  • Could you share some examples of how companies could construct their solutions to use realistic models that could also be fast? 

We work with the concept of fast and slow data, with speed not characterizing how soon they arrive in a data pipeline but instead if the data are slow or fast by nature. 

Consider an application with thousands of users. Then each user has unique data describing themselves. We classify this data as fast data to an application because it is used on the fly for each customer journey. The same holds for all data that describes the users’ preferences or interest in learning about what-if scenarios on the economy. 

On the other hand, the forecast of where the economy could be in 30 years (which one need to consider given pension advice) is slow data. Such data should not change every second unless one has done something unwise in the modelling. 

  • Does it mean that data classified as fast or slow get different treatment?  

Instead of performing all calculations simultaneously and risking building slow journeys, fast data is re-calculated much more often than slow data.

Using these classifications of data, Kidbrooke has managed to combine realistic and holistic engineering for a customer and at the same time help the customers build truly interactive customer journeys, either fully digitalized or physical. As a result, OutRank stands out in model quality while being ten times faster than competitors.

  • That’s exciting! Are there any challenges associated with such an approach? 

The data is not black and white for this classification, and we had to make qualitative decisions when creating OutRank. For instance, it is challenging to determine whether the data around the near future is slow or fast. Besides, measuring the impact of these decisions is also a challenge. 

In addition, it is harder to be compliant with the regulations and make the simulations transparent when opting for more complex modelling. Transparency and compliance issues are critical to consider when building fast API:s.

  • Would you say that this method is something that Kidbrooke’s team came up with themselves? 

Not really; we take the technique used by the big banks and insurance companies to keep track of their financials and make it available to everyone. We are committed to our vision of making informed financial decision-making accessible to everyone at all levels of our service. 

Similar Articles

Trends Taking the Wealth Management Industry by Storm Q4 2023

  • December
  • 2023
In the dynamic realm of wealth management, 2023 has unfolded as a pivotal chapte

Exploring ALM in Life Insurance: Strategy, Risks, and Innovations

  • November
  • 2023
In this interview, Hans Sterte, a seasoned economist, and Senior Partner at Hous

Key Trends in Wealth Management Q3 - 2023

  • September
  • 2023
How can wealth management companies ensure profitable growth during this volatil