Unfortunately, the unstable economic situation is making it increasingly difficult to maintain a profitable retail business. Retailers must be able to predict the future with some accuracy in order to run a profitable business. Forecasting sales and demand are therefore becoming two key aspects of business planning.
Sales prediction or demand prediction – which to choose?
The terms sales prediction and demand prediction are sometimes used interchangeably. However, there is a fundamental difference between them. What does this difference refer to and which prediction should we particularly focus on? That’s what we’ll discuss in today’s article.
To begin with, it is worth taking a moment to recall the relationship between the key terms demand, sales and supply. Demand refers to the amount of products or services that customers would like to purchase in a given period. Sales, on the other hand, is the amount of products or services that were actually sold during that period. For sales to occur, there must be a supply of products or services capable of meeting demand. This is because supply is the amount of products and services supplied that are available during a given period. Therefore, there are no sales when there is no demand. However, there are also no sales when there is demand and not enough supply. Generally, therefore, we can deal with three situations:
- Demand = supply
Ideal situation: customers are satisfied with the ability to meet their needs, and the company is satisfied because it sells all available inventory.
- Demand > supply
Not all customers are able to satisfy their needs, while the company bears the cost of lost potential sales. Such a situation arises, for example, when there is a shortage of a particular commodity in the warehouse or on the store shelf at the time when the consumer would like to purchase it. In a competitive market, the customer can then buy a substitute product/service from a competitor.
- Demand < supply
An unfavorable situation for a company that has frozen money in merchandise lingering on the shelves, loses the ability to use store space and logistical resources to supply products in demand, and runs the risk of losing the value of the product altogether (e.g., as a result of exceeding the expiration date).
Accurate demand prediction avoids situations 2 and 3, or at least minimizes their scale and associated costs. At the same time, we can identify 5 areas where demand prediction brings benefits.
Benefits of demand forecasting
- Optimization of production and inventory
With accurate demand prediction, a company can better predict how much product it will need for a given period This allows it to optimize production processes and control inventory levels.
- Increase sales
Ensuring the right amount of products in stock allows the company to increase its sales and customer satisfaction.
- Better planning of marketing campaigns
By having an accurate prediction of demand, a company can better plan which products (or product categories) and during what period it pays to promote.
- Optimization of prices
With demand prediction and knowledge of inventory, a company can optimize the price of a product to balance demand with supply and maximize profit.
- Cost reduction
With accurate demand prediction, a company can avoid the costs of excess inventory and unnecessary logistics costs.
Sales prediction vs. demand prediction – differences
However, what if we prepare a sales prediction instead of a demand prediction? In such a situation, we risk underestimating. As we have already noted, sales occur when demand meets supply. In a situation where supply is insufficient (lack of goods) then demand will not be met and sales will be lower than they could be. In the extreme case with a total lack of goods on the shelf, sales will be 0. A predictive sales model can correctly predict the lack of sales in such a case. However, using such a model to decide on the right product inventory will result in underestimation and loss of potential sales. To make matters worse, the accuracy rates of such a model can be very high. This is because we may be dealing with a self-fulfilling prophecy:
No goods → zero sales → model predicts no sales in the next period →
decision to not supply the product (since no sales are assumed) → no goods.
And the circle closes.
This is a potentially costly mistake at the model conception stage and a trap into which companies sometimes fall. Meanwhile, machine learning methods make it possible to build and train predictive models capable of predicting demand (and not just sales). Such models take into account a number of different factors influencing demand (including seasonality, price, weather, promotions) and can operate at any level of aggregation (product group/single product, region/store group/single store, etc.).
Accurate demand prediction is the key to success. It allows you to reduce costs, increase sales and improve customer satisfaction. However, these benefits can only be provided by the right selection of data science methods suitable for solving this kind of problem.