Planning and Forecasting Using Predictive Planning

Businesses are more widely adopting predictive planning, which uses statistical analysis to estimate what’s likely to happen in the future based on your organization’s historical data. This information helps CFOs and their finance teams understand how factors such as sales or expenses may evolve, letting them allocate budgets appropriately and improve investment and cash flow planning. Using predictive planning and forecasting can help CFOs and other business leaders identify potential risks in their forecasts, such as supply shortages or cash shortfalls. This foresight makes it more likely they can avert problems and protect their company’s profits and reputation.

What Is Forecasting in Predictive Planning?

Forecasting using predictive planning, sometimes called predictive forecasting, is the process of analyzing historical data and projecting what’s likely to happen. Predictive planning is how CFOs and finance teams use that information to prepare for the future. Finance teams who do predictive planning lean extensively on time series forecasting, which identifies patterns and trends in data recorded at regular intervals, such as monthly sales numbers or daily inventory stock levels, to extrapolate what could happen next. Analyzing time series data, such as this, is useful for understanding cycles, seasonality, and long-term trends, all of which help create an accurate forecast.

For example, a CFO may want to forecast sales for an upcoming holiday season. If the company has years of historical sales data, time series forecasting can provide an estimate that reflects the seasonal impact. However, the finance team must identify and use the best time series forecasting method for the situation to get the most accurate projection.

If analysts have enough quality data to draw insights and apply the models correctly, forecasting methods used in predictive planning should have a higher degree of accuracy versus other practices, such as a gut feeling or assuming a flat percentage increase year over year. In addition, many organizations choose to further validate their forecasts using software with embedded predictive analytics capabilities, which uses data modeling and machine learning (ML) to uncover relationships in the data set that a person may not see. Validating forecasts with predictive analytics is increasingly considered a standard part of the predictive planning process.

Key Takeaways

• Predictive planning is when finance teams use statistical techniques to identify trends and patterns in historical data to estimate future values, such as sales, expenses, and cash flow, to improve the planning process.
• Resulting forecasts and predictions are only as good as the data that goes into making them, so finance teams should only use clean, relevant, and trusted data.
• There are many time series forecasting methods to choose from to do predictive planning, and analysts must find the most appropriate method and carefully consider which variables are necessary to achieve the most accurate forecast.
• Predictive planning can be applied to a wide range of business use cases, such as forecasting cash flows, product demand, and the return on investment for marketing campaigns.

Predictive Planning and Forecasting Explained

Predictive planning assumes that historical patterns and trends repeat to a degree. Therefore, by analyzing the past, CFOs and finance teams can prepare for what’s likely to come by uncovering insights and creating forecasts that anticipate future outcomes based on current data. The adoption of predictive planning and forecasting is on the rise due to the growing demand to forecast trends reliably across an expanding number of use cases and the increasing volatility and complexity in business. The number of organizations that say they use predictive planning productively went from just 4% in 2020 to 27% in 2022, according to a global survey by market analyst firm BARC of 295 employees who participate in the planning process. Another 17% were deploying it or using prototypes in 2022, the survey found. Companies that can accurately forecast the future are more likely to make informed decisions today and create plans that set them up for success tomorrow.

Say a company wants to forecast next year’s sales, raw material costs, and production capacity requirements to see if it makes sense to invest in new equipment. Several factors influence whether the team’s forecast will be accurate. First, the finance team should have enough data to discover patterns and trends. A general rule of thumb is to have at least twice as much historical data as the length of time you’re forecasting—for example, 24 months of historical data to create a 12-month forecast. The data should also be reliable and clean, meaning free from false, duplicate, or incorrectly formatted data. Typically, predictive planning is done using data from finance, which tends to be well structured and, hopefully, accurate. Predictions are only as good as the data used to form them. Additionally, the financial planning and analysis (FP&A) analyst must identify the right time series forecasting model (often multiple models) given the available data and the question they’re answering. Choosing the wrong variables can point to a poor prediction and, therefore, a bad decision and adding more variables can lead to “overfitting,” where the data model begins to model random noise present in the data.

With so many factors to consider, more finance professionals are turning to predictive planning software and services that help them navigate these decisions and, ultimately, get more accurate forecasts faster. The more precise forecasts are, the better finance teams can plan for the future and wisely allocate budgets. Consider how many factors are at play when creating an annual budget and the significant impact one line item, such as hiring costs, may have. A company could use predictive planning to spot historical trends in a company’s attrition level, assess the likely best- and worst-case scenarios, and consider adjusting the steady-state attrition forecast if the model predicts a significantly different outcome.

Beyond the finance team, cross-functional use of predictive planning and forecasting is increasingly vital to deal with volatility in the economy, workforce, supply chain, and other business drivers. Predictive planning can be used in inventory management, for example, to spot cyclical or seasonal spikes that can put an unexpected strain on working capital or shortages that can slow down production. A procurement manager may use predictive forecasting to estimate raw material costs and decide whether to hedge against a commodity price increase. A customer service team leader might use predictive planning to forecast call volume trends to set their staffing levels. Operational insights such as these impact many areas of a business and help organizations create more precise financial plans.

Nearly half of CFOs say their top priority is building predictive models and gaining the ability to analyze and prepare for different scenarios, according to a PwC survey from August 2022. This foresight allows them to avoid potential risks, such as revenue shortfalls or investing too heavily in a new market that’s unlikely to meet expectations. Building scenario plans based on best- and worst-case forecasts prepares teams for how they’ll respond. Further, companies increasingly use predictive planning software that automatically updates forecasts using an organization’s real-time data, letting finance teams see a disaster or success coming sooner so they can rev up their planned response.

Time Series Forecasting Methods

Time series forecasting is a technique that uses historical data points recorded at regular intervals to predict what will likely happen in the future. Numerous time series forecasting methods or algorithms exist, and finance professionals must identify which will give them the most accurate prediction based on the available data and what they want to accomplish.

Time series forecasting generally studies trends, seasonality, and cycles. Trends reflect the gradual or steady increase or decrease in data patterns over time, typically due to long-term factors—things such as changes in the population, organic growth, or changes in technology. You can often model this either with a linear function or maybe a slow-moving curve function. Seasonality focuses on periodic, regular, and somewhat predictable increases and decreases that occur over time. And when discussing monthly data, seasonality will usually occur within a calendar year. It may be broken down into quarters or natural seasonality, such as holidays. Cycles are patterns of increases and decreases that might not be quite as regular and might last more than a year. In business, this is often due to things such as multiyear business cycles that move slower than a typical seasonality pattern does.

Here are popular methods:

• Single moving average (SMA) calculates the average price of an item over a defined time period and works best with volatile data without trends or seasonality.
• Double moving average (DMA) calculates the moving average and then averages that single moving average. This technique uses both data sets to project expected future behavior and works well with historical data that has a trend but no seasonality.
• Single exponential smoothing (SES) weighs the data, placing the greatest importance on the newest data point, and gradually decreases the weight the older the data becomes. This method helps overcome the limitations of moving averages and percentage change methods and works best with volatile data that doesn’t have a trend or seasonality.
• Double exponential smoothing (DES) performs and repeats the SES method. DES is applicable when data has a trend but no seasonality.
• Damped trend smoothing (DTS) nonseasonal method applies SES twice, but unlike the DES method, the trend component curve is dampened and flattens over time. This technique applies to data that has a trend but no seasonality.
• Seasonal additive calculates the seasonal index for trendless historical data, resulting in a curved forecast that shows seasonal changes and exponentially smoothed values. It’s useful when seasonality doesn’t rise over time.
• Seasonal multiplicative works best with seasonal data that goes up or down, differentiating it from seasonal additive. This method also calculates the seasonal index for trendless historical data.
• Holt-Winters additive creates exponentially smoothed values for the level of the forecast and the trend and adjusts for seasonality. This method works well when neither the trend nor seasonality increases over time.
• Holt-Winters multiplicative applies when trend and seasonality rise over time. Like Holt-Winters additive, Holt-Winters multiplicative creates exponentially smoothed values for the level of the forecast and the trend and adjusts for seasonality.
• Damped trend additive seasonal method projects seasonality, damped trend, and level individually and then combines the data into a linear trend forecast. This technique works best when the data has a trend and seasonality, but seasonal variation is fairly constant.
• Damped trend additive multiplicative method also projects seasonality, damped trend, and level individually and then combines them into a forecast. However, it follows a process made for situations where the seasonal variation increases with time.
• Autoregressive integrated moving average (ARIMA) is a calculation that captures trends for one variable over time and predicts future data points by looking at the difference between values in the series. It’s applied when there’s no seasonality, but separate seasonal ARIMA models (SARIMA) exist.

Forecasting Method Selection and Techniques

Predictive planning helps organizations make critical decisions and prepare for what lies ahead. To do so effectively, FP&A professionals must use the most accurate forecasting method given what they want to accomplish and what data is available. It’s also vital that the data is trustworthy, relevant, and the data set is large enough to get the most precise prediction possible. Size recommendations vary, but one approach is to have at least twice the amount of data as your prediction period.

As seen above under the Time Series Forecasting Methods section, each algorithm has caveats and performs better under specific circumstances. For example, if you want to estimate the future price of raw materials in your manufacturing process by looking at its average historical price over a defined period, SMA works best if there is no trend or seasonality. However, if your data has a trend and no seasonality, you’re more likely to get an accurate forecast with DMA. Data can be deseasonalized, but this adds a complication to your model.

In addition to data availability and the purpose of the forecast, analysts must consider factors such as how accurate the estimate needs to be; the costs of creating the prediction in terms of staff time, data sourcing, and computing resources versus the benefits; and how much time they have to conduct the analysis. Finding the most statistically accurate prediction can be a time-consuming process. You need to identify the relevant forecasting methods, run the numbers for each model against historical values, and then analyze which one would’ve had the least errors and best predictions if used in the past. For example, creating a validation data set with a root mean squared error (RMSE) calculation lets you assess your model against historical data points. The RMSE is essentially the standard deviation of residuals on the validation data set, and the lower the RMSE the better. The forecasting method with the most accurate prediction has data points closest to the regression line, which shows the relationship between two variables—the dependent variables on the y-axis and the independent variables on the x-axis of a graph. The right approach could involve using multiple methods.

Many people prefer to use applications with built-in predictive planning capabilities that automate this process. The professional services organization EY surveyed 1,000 CFOs and senior finance leaders for its EY Global DNA of the CFO Survey and found that technology transformation is the primary way they’ll improve the finance function over the next three years, followed by advanced data analytics, which includes using AI to improve financial tasks. These AI applications run a company’s data through various time series forecasting methods, apply RMSE and standard error criteria, and identify the model with the best fit. The application may also project a best- and worst-case scenario along with the prediction.

Some applications allow for multivariate analysis, letting FP&A professionals compare multiple factors at once to improve financial forecasts and corporate planning. Further, it’s possible to automate these processes so as new data becomes available, forecasts and predictions are updated to give CFOs and finance teams the latest insights at their fingertips.

Predictive Planning and Forecasting Use Cases

Predictive planning is becoming essential as businesses face mounting pressure to grow profits and minimize risks amid constant fluctuations in consumer demand, economic conditions, supplier performance, and other variables. A global survey of 303 senior finance executives from CFO Dive and FTI Consulting finds that improving forecasting accuracy and analytics capabilities are two of the top five strategies they’ll use to improve financial performance in 2023 and beyond. Better forecasts with frequent updates enhance an organization’s ability to plan for different scenarios and adapt quickly.

KCB Group, a financial services holding company, used to take more than 12 weeks to prepare and finalize budgets for all of their branches and business lines. Data was in several places, which was one problem. They also relied on market trends and other external data points during planning to forecast for nonfunded income, such as transaction fees and insufficient fund fees, which added complexity to forecasting. Once KCB Group started using an application with embedded predictive planning tools, it was easier for them to use their own business and external data to spot trends and forecast various scenarios. Ultimately, KCB Group cut their budget cycle time by 60% by making improvements throughout their planning process.

More accurate forecasting also helps businesses predict and rapidly respond to market trends to drive profitable growth. When lululemon decided to focus on growing its business outside of North America, the financial planning and analysis team realized they needed to better anticipate how changes in the world economy and industry trends may affect sales. They started using a more robust planning application with built-in predictive analytics, a sophisticated forecasting technique, to forecast multiple scenarios based on their historical and real-time data to continually update their annual plan. The insights improved lululemon’s financial health and strategy, allowing leaders to make better-informed decisions to expand the brand’s reach.

Forecasting has many other uses to support business and financial needs. For instance, companies can project sales more accurately because predictive forecasting can reduce human bias. Statistically based forecasting removes emotion and projects what’s most likely to happen based on past data, letting sales managers and other leaders plan better. Similarly, forecasting product sales over the next six months can help companies create a plan today to ensure they have enough materials to produce goods to meet anticipated demand.

Finance teams often use predictive planning to forecast medium- to long-term cash flows and give them a better idea of their most likely cash liquidity—a major concern for companies of any size. Having cash on hand gives them the flexibility to seize unexpected opportunities or cover unforeseen expenses. However, figuring out how much cash is available at any time can be challenging. For example, if you’re a supplier who sells goods to customers on credit, cash isn’t immediately available at the point of sale for those items. You need to forecast when customers will pay for those credit sales.

Most finance professionals need more than a day to build a consolidated view of their cash and liquidity, according to IDC’s 2021 Global CFO/Treasury Survey. That creates two problems: First, it hinders their organization’s ability to respond swiftly to unexpected situations, and second, by the time they have a number, it’s probably already outdated. The survey also found fewer than 5% of respondents trust their cash forecasts if they’re more than three months out. Considering the complexity of measuring liquidity and its significant impact on the business, more companies are exploring predictive cash forecasting to get more accurate forecasts quickly.

Finance teams also increasingly use predictive models to swiftly validate their forecasts. Predictive models based on machine learning and advanced data analytics can identify relationships in historical data that an analyst may not see. Think of it as a more sophisticated way to generate predictions and insights, especially when analysts are trying to answer complicated questions with many variables.

Forecasting a city’s population growth, for example, is very challenging. City planners need to consider how many people, on average, move in and out of the municipality annually, how many children are born each year, how many men and women there are, how long they’ll live, and other factors. The more accurately they can anticipate changes in the city’s size, the better they can serve that community by building roads and schools, preparing for water and energy usage fluctuations, and make other vital decisions. Predictive models can help with these kind of predictions.

A potentially lifesaving use of predictive planning is in emergency rooms. Hospital administrators can use predictive analytics to forecast patient volumes and plan appropriate staffing levels. In general, ERs have a four-hour rule, where staff must see, treat, and decide if a patient will be admitted or discharged within that time. A 2022 British study of more than 5 million patients published in the Emergency Medicine Journal found that waiting more than five hours in the ER before being admitted to the hospital increased a patient’s likelihood of death over the next 30 days. In a time when hospitals are dealing with nursing and physician shortages, predictive planning and forecasting offer a valuable tool for deploying employees as effectively as possible.

Plan Better with Predictive Planning and Forecasting

A data-driven approach to forecasting can reduce human bias and lets finance teams quickly identify the most likely outcome across multiple scenarios so CFOs can work together with other leaders to make more-informed decisions. Predictive planning and forecasting through Oracle Cloud Enterprise Performance Management (EPM) Planning, part of Oracle Fusion Cloud Enterprise Performance Management, connects planning across finance and lines of business. Each area benefits from access to prebuilt planning models to quickly explore multiple scenarios. Finance teams can tap these forecasts and data models to make more accurate and informed plans that help companies prepare for best- and worst-case outcomes in ways that protect and profitably grow the business.

Predictive Planning FAQs

What is predictive planning?
Predictive planning uses what we learned from the past to plan for the future. Time series forecasting methods project likely future values, such as sales numbers, stock prices, and monthly expenses, based on the assumption that patterns and trends in the historical data will repeat, and tools such as machine learning and AI can be used to validate those forecasts quickly.

What is predictive forecasting?
Predictive forecasting, more commonly called forecasting, analyzes historical data to estimate what's likely to happen by identifying patterns and trends in data recorded at regular intervals.