Predictive analytics is a method of data analysis used within the financial services industry – and beyond – to forecast business-related outcomes. It sits on a spectrum, beginning with descriptive analytics, the most basic form of data analysis, then moving
to diagnostic analytics, predictive analytics, and finally prescriptive analytics – the most advanced of them all.
In this instalment of Finextra’s Explainer series, we explore how financial institutions (FIs) can achieve predictive analytics, the kinds of data it is powered by, and its various applications.
Descriptive, diagnostic, predictive, and prescriptive
Descriptive analytics is the simplest form of data analysis – enabling FIs to understand exactly what is happening in their business. It involves cleaning, relating, summarising and visualising vital data. This may look like a bar graph showing sales volumes
of a specific product across a calendar year. Such knowledge is foundational to effective resource allocation.
Slightly more complex is diagnostic analytics, which looks to answer why certain trends are occurring. For instance, it might answer: What's causing revenues to decline in the winter months? How do these figures relate to other business lines? By
creating a dashboard using diagnostic data, treasurers can easily make adjustments to ensure any weak spots in cashflow, liquidity, or investments are rectified.
Predictive analytics goes one step further. Using statistical modelling or machine learning (ML), FIs can forecast what will happen next within the business. For example, if in the coming months sales of a particular product are predicted to shoot up, inventory,
human resources, or technological tools can be diverted to that business line. Likewise, if sales are predicted to slump, course corrections can be executed to support specific departments.
Simply put, predictive analytics enables FIs to get innovative with their decisioning. What’s more, the longer the model runs, the more the predicted figures can be compared against the actuals – enabling FIs to gradually increase the precision of the instrument.
It is only once these capabilities are established that FIs can begin building out prescriptive analytics, which essentially advises on pro-active actions, based on the predicted forecasts.
The methods
Statistical models help determine patterns in information – including historical and current data – which are used to make the predictions. Here are several types of methods that can be used to enable predictive analytics:
- Decision trees: These models quickly help FIs understand what leads to a customer’s decisions. As the name implies, it looks like a tree – where the branches signify the choices available and each leaf represents a specific decision. The benefit of a decision
tree is that it is easy to interpret.
- Regression: Perhaps the most common of all statistical models, and particularly useful for identifying patterns in large data sets, regression models can help FIs understand, for instance, how price is shaping the performance of a stock.
- ML: Increasingly common is the use of
neural networks, which work in a manner similar to that of the human brain. Once again, thanks to its advanced pattern recognition capabilities, ML – as underpinned by neural networks – is ideal for working with data that has complex relationships. It can
also be leveraged to confirm the findings of the previously listed models.
Other options for FIs include cluster models, which aggregates data with similar attributes; or time series modelling, which assesses inputs at given frequencies – be they daily, weekly, or monthly.
The applications
The exact methods deployed may vary, but the findings of predictive analytics can be applied to numerous areas within a bank. Here are some examples:
- Risk management: Predictive analytics can identify and mitigate risks by forecasting potential loan defaults, fraud instances, or market downturns.
- Fraud detection: Particularly effective here is ML, which can digest vast datasets (such as payment history); spot patterns or anomalies in transactions; and identify instances of fraud – sometimes before the crime even occurs.
- Customer segmentation: By analysing historic consumer behaviour or preferences, products and services can be personalised to drive user satisfaction. Market demand and behaviour can also be leveraged to inform pricing strategy.
- Treasury: There are many uses for predictive analytics within the treasury department. Whether it be cash flow forecasting to better manage liquidity, or market trend forecasting to optimise investments, predictive analytics enables treasurers to get pro-active
and embed themselves as a driving force in business strategy.
The bottom line
The aim of predictive analytics is to forecast business events and improve decision-making processes. Ultimately, this can be exploited by FIs to keep costs down as well as drive revenues. Often, robust data analysis is the difference between the success
and failure of a given business line.
Despite all the benefits, FIs must remain cautious when handling data. In some jurisdictions, the use of predictive analytics has been restricted, due to its potential to generate unfair outcomes. This can be the refusal of loans to certain ethnic groups,
discriminatory credit scoring, or refusal of employment. An example of this came in September 2023, when Rhode Island bank agreed to pay
$9m over redlining allegations.
Providing the data being used is organised, clean, and representative, predictive analytics can be transformative for all kinds of institutions. Treasurers already with dashboards in place should consider moving to develop prescriptive analytics – the holy
grail of data analysis.