The Concept of Demand Forecasting


The Concept of Demand Forecasting

Whether you’re deciding how much of a given product you should order for the upcoming season or which products should feature in your next marketing campaign, understanding the current and future demand for your products makes your decisions that much easier. Rather than stabbing in the dark and risking reduced profits or stock-outs, Demand Forecasting enables you to reduce the uncertainty and risks commonly associated with decision-making.


We’ve written about Demand Forecasting before (it’s a topic that we’re pretty passionate about), but let’s break it down even further, let’s get conceptual.


For those who are unfamiliar with the term, Demand Forecasting describes a suite of analytical tools that aim to predict the demand for products and services for a certain period of time. If you need a deeper dive into the topic before continuing, head here (then come back). Or, if you feel you’re abreast of the subject, read on to find out about the components that make up Demand Forecasting, the kinds of data that you’ll need, and who Demand Forecasting is best-suited for.


Why do we need to forecast demand?


In the same way you use a weather forecast to decide whether (forgive the word play) you’ll plan a weekend beach-trip or grab an extra jacket on your way out the door, businesses use demand forecasts to make all kinds of decisions in both the long and short term. These forecasts rely on data from a range of sources, from expert opinions and historical sales data, to weather patterns, to upcoming events and holidays, and can be chosen and customised based on the needs and goals of your business.


Namely, the best type of forecasting to use depends on:


  • The time period you want to forecast for

  • The variables you want to consider in your forecast

  • The products you’re forecasting for

  • Data availability


These factors are crucial in creating a forecast with the greatest accuracy and make up the core components of a Demand Forecasting method. Since they are vital for choosing the optimal forecasting method, let’s look at them in greater detail.


Length of Forecasts


The length of a forecast refers to how far in advance the forecast will be predicting demand for. Typically, the forecast is split into three categories: short, medium, and long-term. While the specific time period for each category isn’t well-defined, in a retail and manufacturing context, short- term forecasting anywhere between 0-3 months, medium term is 3-6 months, and longer term is 6+.


The length of forecasting you opt to use is crucial, and it’s not a case of one-size-fits-all. Different term lengths are better suited for different methods of demand forecasting and for achieving your specific goals and purposes.


Starting small, short-term forecasting can be used to make tactical and operational decisions. Short-term forecasts are often broken down further into weekly or daily forecasts, depending on the operational cadence within your business. In traditional businesses, quarterly forecasts are used to budget and plan sales, while monthly and weekly forecasts can be used for short-term capacity and inventory planning. On the daily and hourly level, these kinds of forecasts are often used for inventory deployment and planning of transportation and production.


Medium and long-term forecasts, on the other hand, are used for broader strategic decisions. These forecasts assist planning of sales, marketing, and capacity, or investment strategies - think machine replacement schedules and decisions involving production capacity.


Forecast Scope


The scope of your forecast refers to the types of variables that it takes into account, and is generally divided into internal and external Demand Forecasting. Internal forecasting, as you might guess, solely considers current business practices and takes into account variables such as:

  • Revenue

  • Profit margins

  • Cash flow

  • Product costs


Conversely, external (or macro) Demand Forecasting considers market conditions and external environmental factors when predicting demand for a company’s product or service. This level of forecasting is often used to drive internal business decisions, such as those relating to expansion and development of customer segments, and evaluating product portfolios.


Product Types


While it might be surprising to some, the types of products you want to predict demand for can be a pretty important determiner of successful forecasting. Generally, your products will fall into one (or more) categories:


  • Capital goods - machinery, tools, raw materials, and even factories themselves.

  • Durable consumer goods - these products generally have ‘low-velocity’ demand, can be used more than once, and include everything from cars and furniture to electronics and clothing.

  • Non-durable (perishable) consumer goods - from food to medicine, this category includes anything that is consumed once.


The demand for capital goods is derived, meaning that it depends on the demand for the consumer goods they are used to produce, as well as the rate of industry growth, market size, and the level of capacity utilisation.

On the other hand, consumer goods - both durable and perishable - have direct demand since these products are meant for consumption by end-users. For durable goods, demand is influenced by their obsolescence rate and maintenance costs, as well as socioeconomic variables, whereas the demand for non-durable goods is largely affected by the purchasing power, price elasticity, and other demographic variables of your customers.


What kind of data do you need?


To undertake Demand Forecasting, you’ll need data. While that might seem like a huge understatement (it is), the quality, quantity, and types of data you have at your disposal is the main driver in determining which type of Demand Forecasting method is right for you. Broadly, Demand Forecasting methods use different types of data depending on whether they are qualitative or quantitative forecasting methods.

To start with, qualitative methods tend to focus on the wider economic climate, and rely on sources of data such as expert opinions and market research. These methods are best suited for forecasting products with little to no historical sales data and for long-term forecasting. Conversely, quantitative methods rely on hard data - think historical sales, financial reports, revenue figure, and website analytics. This data is used to create statistical models and perform trend analysis so that a forecast can be produced. The more powerful quantitative methods employ Artificial Intelligence to improve accuracy, uncover patterns that would be undetected in manual or statistical analysis, and produce lots of forecasts quickly. Want to figure out which method is right for you? Head here to read about the different Types of Demand Forecasting at your disposal.


For instance, Machine Learning time series methods are quantitative methods, but can use data from a wide range of sources as input (known as Multivariate). Although historical sales data is the an important input, Machine Learning methods can gain vital accuracy by ingesting additional data streams - think price, promotions, weather, and, in the current climate COVID-19 data from John Hopkins University- as well as metadata such as brand, product category, and product specifications.

Who uses Demand Forecasting?

If you’re in the business of producing, selling, or supplying a product or service (also defined as being in business), Demand Forecasting can help you.


Demand Forecasting is especially useful for retail and ecommerce, since accurate demand forecasting lead to improved inventory management, optimised stock levels and restocking schedules. And, to find out how Demand Forecasting works for start-ups in particular, head here.


Common Issues and Solutions


One of the issues that can arise with Demand Forecasting is information overload. Basically, when you are forecasting in the short-term or for a large number of products, it can be difficult to sift through that much information. Luckily, Demand Forecasting platforms (such as our very own) can do all of that hard work for you. Machine Learning algorithms are able to manage lots of forecasts, even down to a specific product or product-per-location (AKA SKU and SKUL) can help to reduce stock-outs and overall inventory costs by up to 15%.


Another issue relates to forecast errors. For many, forecasting isn’t adopted because of a belief that your business is unique and unpredictable, or it was tried once with uninspiring results. Given the improvements in the space recently, if you were once bitten, it is no longer time to be shy.


Problems in forecasting for products with intermittent or sporadic demand, aka ‘low velocity’ products? These errors can result in reduced forecast accuracy, and often arise when forecasting strategies:


  • Don’t account for seasonal demand

  • Use identical forecasting calculations for products that behave differently (in terms of sales)

  • Don’t filter out promotional sales activity, which can skew actual demand trends


Solution? Our Low Velocity suite of Demand Forecasting tools that treat intermittent demand differently (I can’t go into the how just here - sorry!)



Final Thoughts


Adopting Demand Forecasting into your business strategy can have significant benefits when it comes to inventory management, mitigating risk, and, most importantly, improving revenue.


Plus, software such as Remi’s Demand Forecasting platform can make the quest for a Demand Forecasting method that suits your needs simpler (while also improving your forecast accuracy).


Our Demand Forecasting suite uses an Ensemble Method - a combination of different Machine Learning models that run simulations until a consensus on the best result for your products is reached. This, alongside the watchful eyes of our Quality Assurance Team, ensures that your forecasts are:


  1. Accurate - our team audits the accuracy of models regularly to mitigate the risks that come with forecasting

  2. Suitable for your business


Want to find out more about demand forecasting? Why not have a read through of our case studies, and find out how we’ve used demand forecasting to help increase stock availability and improve the accuracy of forecasts. Once you’ve had a look through those, why not check out our blog for the latest AI reads. Or, if you’re looking to implement Demand Forecasting for your business or improve your current forecasting methods, drop us a line here.