In this guide, we will answer a few of the key questions around the value that systematic forecasting can offer to enterprise companies, and help you decide if it is something that could benefit your team.
Why forecast at all?
Forecasting can take different forms at different businesses, and it is important because it enables companies to be proactive, rather than reactive, in planning for the future. As informed guesses about the future, forecasts help companies to agree achievable goals in the short, medium and long term. Forecasting also allows teams to decide how to allocate their resources, whether time, money or something else, over given time periods in order to meet their goals. Forecasting can also provide a competitive advantage by enabling a company to anticipate changes that might affect your business before its rivals.
What different types of forecasting a business can use?
In general, forecasting falls within two main fields; qualitative and quantitative. Qualitative forecasting relies on compiling expert knowledge, rather than data analysis, to predict a future outcome. This type of forecasting may take different forms in different industries, but in commodity markets usually will rely on the opinions of market analysts, experienced traders, category managers or specialist consultants. Qualitative forecasting can be very valuable in reacting to short-term market uncertainty where past trends cannot be relied upon to predict future outcomes. This is particularly true following ‘black swan’ events which by their definition are extremely rare and unpredictable.
The shortcomings for qualitative methods mostly result from a form of bias; whether towards recency, where experts place too much emphasis on the impact of recent events in evaluating future trends, or based on their own biases regarding the main driving forces of specific markets.
Quantitative forecasting, on the other hand, relies on data analysis rather than expert opinion. There are different niches to approaching this method, but generally quantitative methods are built on identifying patterns in historical data and using that information to forecast future trends with algorithmic models. Quantitative methods can be very effective in identifying market forces and enable companies to implement strategies based on objective data rather than subjective opinion, allowing for more visibility into decision-making. Depending on the sophistication of the model used, however, quantitative methods can be expensive for companies to implement and can also suffer from difficulty in analysing the impact of unusual market drivers.
What value does AI and Machine Learning bring to forecasting?
There are several key advantages of using AI and Machine Learning methods in forecasting.
- Computers can collect and process data faster than people. Given that quantitative forecasting methods rely on both the sample size and relevance of the data used, the ability to save hours of employee research when selecting and processing data is a key benefit for organisations.
- Beyond importing the data, machine learning can analyse thousands of data points at once and identify patterns or trends with much greater sophistication than people, reducing the burden on employees to conduct analysis and increasing business efficiency. For example, analysing weather patterns over a particular growing region for a crop over the past decade to identify correlations with the crop price would be almost impossible for a person, but can be completed quickly by an algorithmic model.
- Machine learning models improve over time. Ongoing research into adjusting the training horizons for the models, importing new data sources or investigating incidents where the models were incorrect all allows an increase in accuracy over time. While neither quantitative nor qualitative forecasting will ever be perfect, it is possible to make incremental improvements over time with Machine Learning methods.
How Accurate is Systematic Forecasting?
To assess the value of forecasting, particularly quantitative methods, there are numerous accuracy metrics that can be considered. Indeed, evaluating the same data with two different accuracy metrics can often lead to contrasting performance results. Therefore, rather than suggesting a particular metric that can be used to assess every forecast, below are some suggestions of points to consider when evaluating the accuracy of a forecast and the value it can therefore add to your team.
- Forecasts should be evaluated based on the value they contribute to business performance, rather than in isolation. Forecasting helps companies to achieve goals, whether that is in budgeting, negotiating or implementing long-term strategies, and therefore how much a forecast can support those outcomes is the key indicator of its value to a company. Can you measure the value of a forecast in efficiency, profitability or another business-related metric?
- The timeline of your goals will significantly influence how you should interpret a forecast to gain the most value from it towards achieving those goals. Assessing a 1 week forecast by the same metrics as a 12 month forecast will inevitably make the performance of one seem poor.
- Forecasting, regardless of the chosen method, should not be considered as a crystal ball but rather as part of a range of tools that can be used to secure better business outcomes in the future. Acting on forecasted information might not always lead to your desired results, so it is vital to understand the value you are seeking from a forecast before acting on it.
