4 ways CFOs can maximise the benefits of predictive analyticsAchieving maximum value from predictive analytics requires a robust strategy and a model with a human-centric design that is also appropriate to the business.
Whether it's data mining, predictive modelling, or a blend of both, companies increasingly rely on data analytics to pinpoint relevant data patterns, trends, and projections to scale up commercial operations and increase profitability.
As big data proliferates, companies (and the data scientists they employ) are using predictive analytics technologies to forecast future cash flows, decrease costs, reduce the likelihood of fraud, and even improve employee welfare by eliminating some manual tasks in their organisation.
What is predictive analytics?
Predictive analytics attempts to forecast future events using techniques that include data mining, statistics, modelling, AI, and machine learning to analyse current and historic data.
This type of analytics is often used to answer questions about customer behaviour, identify new opportunities, and suggest how best to target resources for maximum return.
The goal for any CFO using predictive analytics is to learn from their existing data how to predict what will happen in the future.
There are various applications of predictive analytics that include forecasting, strategic planning, performance management, and identifying risk patterns.
Predictive analytics can do this in a variety of ways using different models (see the sidebar "Predictive Analytics Models").
Using predictive analytics as a CFO
Using predictive analytics to drive profitability in a commercial enterprise is not new. The challenge for CFOs is to maximise the benefits of data analytics.
Ash Noah, CPA, FCMA, CGMA, vice-president and managing director–Learning, Education & Development at AICPA & CIMA, together as the Association of International Certified Professional Accountants, spoke to FM, along with Deloitte's US-based Benjamin Booher, senior manager, Tax Technology Consulting.
They identified four tips for CFOs to achieve maximum value from using predictive analytics.
Build a robust predictive analytics strategy
Both Noah and Booher agreed that the starting point in using predictive analytics is for finance leaders to consider a robust predictive analytics strategy.
Even with good design and meticulous planning, a predictive analytics strategy only yields best results when the right leaders are in place. Booher explained that companies need an "executive leadership support team" to ensure full adoption and maximise the value of their predictive models.
Use analytics to forecast sales
Finance leaders are increasingly relying on predictive analytics to forecast sales. It is an important element to gain a competitive advantage and stay profitable.
"Regardless of the specific application, an organisation should approach the process of AI-enabled forecasting with both a short-term focus to mitigate uncertainty in the current cycle and a long-term focus where planning processes are enhanced," Booher said.
He added: "This will develop more dynamic and effective forecasting capabilities."
"[Predictive analytics] work best on pinpointing sales flow. Maximising revenue through these programs will give you a greater rate of return than devoting your resources to a model that helps you cut a few costs," Noah explained.
Smaller enterprises can use readily available analytics software for predicting sales.
"If you take a restaurant as an example, less advanced software can still analyse orders in the kitchen and identify the busiest time of the week to give [the business] actionable solutions and insights remotely with little cost," Noah said.
Choose an appropriate predictive analytics model
A good forecasting model can be the difference between achieving record profits or experiencing an insufficient return on investment (ROI). Technology shortens the time for strategic decision-making so you can execute straight to market.
Noah pinpoints hotels as an example.
He said: "Look at the hotel chains. Their prices change in seconds based on predictive demand. Hotels change their prices, like airlines, through predictive analytics to achieve appropriate pricing methods that practically guarantee sales. Imagine a person trying to do that?"
Booher stressed the importance of selecting the forecasting model appropriate for the business.
"Many organisations attempt to deploy out-of-the-box algorithmic models that do not align with their business needs," he said.
While out-of-the-box models start out less costly, they can become expensive once you start customising your technological requirements. Out-of-the-box solutions are unlikely to yield the same success as models directed to a particular business need. "These [out-of-the-box] models are less likely to be adopted and, as a result, may fail to achieve the anticipated revenue," Booher suggested.
Ensure the model has a human-centric design
Technology is constantly evolving, so expecting the workforce to stay one step ahead isn't always realistic. A user-friendly model, especially one that is not overly prescriptive, is critical for effective user adoption.
"Whilst every organisation has a different perspective on ROI, we believe several factors are critical to success with predictive analytics," Booher said. "[These] include an effective end-to-end strategy and design. To enable an automated forecasting environment, design efforts should span functions, data sources/systems, tools, processes, and talent."
A user-friendly, human-centric design, according to Booher, "focuses on the end user to better align forecast capabilities with related processes and outputs".
He said: "Many organisations fail to maximise the value of their predictive models because they are viewed as black boxes where results cannot be fully explained and, as a result, are not fully incorporated by finance leaders."
Using these steps can avoid this adoption pitfall and maximise value for the business.
— Hugo Johnson-Driscoll is a content writer at AICPA & CIMA, together as the Association of International Certified Professional Accountants. To comment on this article or to suggest an idea for another article, contact Oliver Rowe at Oliver.Rowe@aicpa-cima.com.
Predictive analytics models
Predictive models and techniques used by companies include:
- Predictive forecasting: Predictive forecasting assesses a company's data and estimates financial outcomes. The insurance industry, for example, assesses the risk of an individual or group by analysing data such as age, gender, health status, driving record, and more before deciding on the appropriate premium.
- Time series analysis: Analysing datapoints that have been collected over time — time series analysis — can show how variables change. For example, it can provide information on customer or visitor numbers and seasonal variables relating to purchase points or the number of purchases. Companies, including Coca-Cola, use sophisticated weather modelling to predict their most profitable (or least profitable) quarters and estimate harvest outputs for their raw ingredients.
- Classification model: This is a more basic predictive analytics model placing data into categories based on historical data and can also be used where a binary output is required. In banking, classification models predict stock and share price movements.
- Outlier model: Analysing notable dataset deviations, outlier models can help detect future fraud. For example, they can link a customer's credit card to a purchase in a city without any previous transaction history there.
Data Analytics Core Concepts Certificate