Artificial Intelligence
What is demand forecasting and why is it important?
People have been attempting to predict the future for thousands of years, with varying levels of success. Consider this quote from the chairman of IBM in 1943, “I think there is a world market for maybe five computers”. In AD357, the ruler of Rome, emperor Constantine, declared it forbidden to participate in forecasting. We are constantly inundated by people who claim to be able to tell us what stocks are going to rise, when the next major war will occur, and even when the world will end.
Thankfully, the world hasn’t ended yet, but major events in the market can certainly cause great amounts of panic and erratic collective societal behavior. This has become more relevant now more than ever with the rise of the COVID-19 pandemic, leading to mass unemployment and highly unpredictable market movements. Luckily for us, over hundreds of years, people have helped develop a set of tools that allow us to mathematically predict certain behaviors.
Demand forecasting is the art of using historic information, such as past sales or stock market data, to help get a good idea of what the future will look like. This task is fundamental, crucially important to running a business smoothly and making sound operational decisions, and notoriously difficult to perform accurately. Demand forecasting is so pivotal because it allows a business to set correct inventory levels, price their products correctly, and understand how to expand or contract their future operations. Poor forecasting can lead to lost sales, depleted inventory, unhappy customers, and millions in lost revenue.
How is demand forecasting done?
There is a seemingly endless number of ways to predict future demand of a product or service, both qualitative and quantitative. One of the simplest methods used in industry is to say something like “well my sales at this time last year were X, so I think this year they will be about X”. This can quickly spiral out of control, however, into things such as “but my competitor introduced a similar product this year, so that could reduce X to X-Y, but we also saw the economy and population of the service region expand, so that could increase sales to X-Y+Z,…”. Using simplistic techniques can sometimes yield the best results, often in situations where sales are very erratic, but many times they gloss over important information which could help improve prediction accuracy. You may have picked up on some of the important factors to consider in a good demand forecast in the example above, such as geography, competition, and economic indicators, but other things can come into play as well, such the stock market, seasonal trends, overall trends, causal relationships to various other products, and even weather.
How can I improve my forecasting?
The best way to optimize a forecast is to truly understand the driving business processes behind an industry. This knowledge can help determine what factors will be critical to include in a prediction. For example, in an industry that sells travel packages to the beach, we might expect weather to play a role in how many people decide to take a trip to the ocean. Once the problem domain is understood and a solid picture has been developed around the root processes involved, rather than using a qualitative approach to predict demand, many large companies are moving toward more advanced techniques, such as automated machine learning and algorithmic predictions.
Using computers and machine learning to predict demand has many benefits. A few of these include the ability to automatically predict the behavior of large portfolios, removing human bias, and using statistics to predict something that is inherently statistical in nature. A mathematical model is able to accurately and quickly extract information from data that would not be readily evident to a human. Artificial Intelligence has become prevalent in the demand forecasting world, allowing companies to develop deep insights around what is driving their sales and what their sales will look like in the future, by doing things like finding leading sales indicators, factoring in mergers, major market events like Covid-19, and helping determine how well a new product will perform in various markets.
Advanced algorithmic modeling can often help improve forecasting by including multiple variables and external information to help a computer learn how different factors in the market will impact sales. Model factories, algorithms which test many different models for every product, can even further optimize forecasting by building the best possible forecast for each product rather than dealing with the “one size fits all” types of solutions that are commonly used.
What are the next steps?
Developing an advanced understanding of how to leverage machine learning and cutting edge forecasting techniques as well as determining the driving factors and underlying processes of an industry will push a company to the next level of competition, delivering not only a better and more efficient business but also providing a critical leg up over the competition. The famous forecasting expert Paul Saffo said, “The goal of forecasting is not to predict the future but to tell you what you need to know to take meaningful action in the present”. Even small improvements in product portfolios can lead to tens of millions in revenue generation, so make sure your business is taking the right steps to grow and prosper.
Aimpoint Digital will help you take an idea from thought through execution. This collaborative journey will enable you to get the most out of your data and technology investments. Contact us to begin your acceleration process.
Author
