🇺🇸
A sales forecast estimates how much you will sell over a future period, based on your historical data and the variables that move your market. It is also called a sales projection or a sales revenue forecast, and it is the basis for every purchasing and inventory decision downstream.
The sales forecast is an early estimate of the future revenue of a product or service. It is based on the analysis of historical data, studies of customer behavior and market trends.
Companies in their tactical operation are constantly reviewing the sales forecast as a starting point for defining strategies to achieve growth objectives; analyzing how to optimize resources in production, logistics and marketing strategies.
These three terms get mixed up in the same meeting and they are not the same thing. The practical difference is which method you use and which decision each one supports.
Past records of sales and buying patterns that help to understand the behavior of the market where trends are found over a period of time.
Period of time to be analyzed; this can be in months, quarters, semesters or years.
Changes that occur in the market that may influence sales, such as macroeconomic variations, technological advances, and customer and competitive preferences.
It is the analysis of information about potential customers, market segments, brand perceptions and external factors that could affect sales.
Elements outside the company's control but that may affect sales, such as global economic conditions, government regulations, changes in weather and unforeseen events.
Consumer habits, preferences, needs and motivations that influence their purchasing decisions.
Study of competitors' marketing strategies, prices, products and services to evaluate how they may affect sales and anticipate possible competitive actions.
Feedback and information provided by the sales team, account managers and other departments within the organization that are related to business management and can provide valuable information about market trends and customer expectations.
To build a sales forecast properly, follow this process:
There are different types of sales forecasts; the one you want to use depends on the industry and the source data you have. I'm going to show you the options you have and then we're going to look at several examples.
There are different types of sales forecasts; the one you want to use depends on the industry, the data you have available and the tools.
The moving average makes it possible to calculate the future value based on the average of previous data over a specific period of time. It's as if we had a window that moves through our data, and at each moment, we calculate the average of the values that are inside that window.
The choice of the type of moving average and the length of the period will depend on the data and the purpose of the analysis. Longer periods will soften the data more, while shorter periods will be more sensitive to short-term fluctuations.
For example: Using a 3-month moving average, the forecast for month 7 would be: (100 + 110 + 115)/3 = 108.33 (thousands of units)
Statistical relationships are established between the variables that affect sales, such as price, advertising, consumer income, etc.
For example:
Using a simple linear regression, we obtain the equation: Sales = 70 + 10 Month The forecast for month 7 would be: Sales Month 7 = 70 + 10 7 = 140 (thousands of units).
With the advent of artificial intelligence, predictions can be made through machine learning and deep learning; forming millions of scenarios; and entering variables that a traditional statistical model cannot read.
For example, at Datup we use intelligent machine learning and deep learning models to forecast demand, where we enter external variables that directly affect each industry and company, combining both quantitative and qualitative data.
A group of industry experts meets to discuss and deliberate on market conditions and emerging trends. Individual opinions are collected and combined to form a consensus forecast.
For example, after considering external factors, such as a change in sugar regulation, you decide to adjust the forecast downward by 20%.
Similar to the expert panel, the Delphi method involves gathering and iterating expert opinions anonymously. A series of rounds of questions and feedback are conducted until a consensus is reached on the forecast.
The Delphi method takes advantage of the collective knowledge of experts, reduces the influence of dominant personalities and allows all opinions to be considered equally. However, depending on the selection of qualified experts, it can be time consuming and results may be biased if the panel is not representative or there are extreme opinions.
Computational models are used to simulate different scenarios and evaluate their impact on sales. Various market conditions and business strategies can be explored to identify the best options for the company.
Analytical and statistical techniques used to predict future sales fall into three families: qualitative, quantitative and machine learning.
They rely on the opinion of industry experts to forecast future sales. Experts use their experience and knowledge of the market to make subjective estimates of demand, considering factors such as changes in market trends, economic conditions and competition. Qualitative methods include the Delphi method, opinion polls and scenario analysis.
These have to do with the analysis of numerical data and statistical techniques to forecast future sales. They use historical sales data and other quantifiable factors to develop statistical models that identify patterns and trends in the data. Common quantitative methods include time series analysis, regression models, and exponential smoothing techniques.
Computational algorithms and predictive models are used to analyze large sets of sales data and predict future demand. Machine learning and deep learning models can identify complex patterns in data and be continuously adjusted to improve forecast accuracy as more information is collected.
The method you decide to apply will depend on the type of industry. To stay ahead and improve productivity, we recommend artificial intelligence techniques, which combine qualitative and quantitative information more efficiently and with better results.
It is the average of the absolute differences between actual and expected sales over a given period of time. It provides a measure of the overall accuracy of the forecast, where a lower value indicates greater accuracy.
This metric calculates the average of the absolute percentage errors between actual and expected sales. It provides a measure of the relative accuracy of the forecast, expressed as a percentage of the actual value. It is useful for evaluating the accuracy of the forecast in relation to the size of sales.
This is the most commonly used indicator, since it includes sales weights according to the SKU or business unit. It is the weighted average of the absolute percentage errors between the predicted values and the actual values.
Bias indicates whether the forecast tends to consistently overestimate or underestimate actual sales. A positive bias means that sales are overestimated, while a negative bias indicates an understatement.
One of the most common tools is Excel. The analysis that is carried out is statistical with linear regression methods, time series, among others. These functions allow users to perform advanced statistical analysis on sales data sets to identify patterns and trends that can aid in forecasting.
It's a data visualization platform that allows companies to intuitively analyze their sales information and generate sales forecasts by identifying trends and patterns in historical data.
Microsoft's data visualization software, which lets you build visualizations and sales projections from historical data connected to your operation. Like Tableau, you can read sales information intuitively and in real time.
Tool as a Service (SaaS) that is located in the cloud; which performs demand and inventory forecasts with advanced models with artificial intelligence and machine learning. Some of the algorithms used are neural networks, which analyze different scenarios, resulting in the forecast with the least error after the iterative analysis.
See more about sales forecasting software for your operation.
Sales forecasting is one of the most strategic processes in any company, because critical decisions in several areas depend on it:
A company that does not estimate its sales ends up buying too much and holding stock that does not move, or running out and losing sales. The value of the forecast is working with future data instead of reacting to past data.
In our experience, we have had different benefits in some areas of the company; I'll tell you more about it below:
By anticipating future demand, companies can manage their inventories more effectively, avoiding both shortages and excess stock. This helps reduce the costs associated with storage and inventory holding, with a direct impact on cash flow.
Sales forecasts provide valuable information that supports fast decision-making across every area involved, from production planning to purchasing and distribution. This helps minimize risk and take advantage of growth opportunities.
By anticipating sales, companies can make sure they have enough product available to cover their distribution channels. This improves customer satisfaction by guaranteeing a positive buying experience and avoiding lost sales due to stockouts.
We have found low assertiveness in forecasts, because they do not have enough historical data to read, making it difficult to learn intelligent models over long time intervals.
Previously it was a topic that was not thought of, now with climate change, war, and currency variations, it is becoming increasingly important to consider these variables in sales forecasts. As well as company-specific information, applicable holidays, business discounts, and more.
Do not update the forecast model with actual sales data as it is generated. It's crucial to incorporate the latest information to improve model accuracy and adapt to changes in sales patterns.
Use a single forecasting method, such as moving average or linear regression, without considering other techniques that might be more suitable for the specific data set. It is advisable to try multiple approaches and compare them to select the one that provides the most accurate results.
Treat all product categories or geographic regions as a single entity, rather than creating specific forecasts for each segment. Different products or regions may have different sales patterns, so it's important to adapt forecasting models to each segment to capture these differences.
Don't take into account seasonal patterns in sales, such as peaks during certain times of the year (for example, Christmas or summer vacation). These patterns must be considered in the forecasting model to avoid significant underestimates or overestimates.
Failure to establish clear metrics to evaluate the accuracy and performance of the forecasting model. It is essential to regularly compare forecasts with actual sales and to calculate metrics such as the average absolute percentage error (MAPE) to identify areas for improvement and adjust the model as needed.
In conclusion, forecasting sales offers significant benefits for companies. Primarily in:
By anticipating future sales, companies can take proactive and non-reactive actions, allocate resources efficiently, and devise strategies that align the areas involved to meet the corporate objective.
Having advanced tools and smarter models will be essential to improve assertiveness and be a competitive company in the market from the supply chain.
In practice they are used interchangeably. The useful distinction is the method: a sales projection extends an observed trend with a simple formula, while a sales forecast models demand behaviour using internal and external variables. A sales budget is neither — it is the target the commercial team negotiates.
With three functions: AVERAGE for a moving average on stable data, TREND for a linear regression, and FORECAST.ETS for series with seasonality (Excel 2016+). The table in the tools section has the syntax.
There is no single one; it depends on whether your data has trend, seasonality or neither. The simplest is the growth rate: previous period sales × (1 + rate). With three months of 100, 110 and 115 thousand units, the moving average gives (100 + 110 + 115) / 3 = 108.33 thousand units.
Using the history of an analogous product, market research or an expert panel. With less than 12 months of history no statistical or machine learning model can learn seasonality. We break it down in how to estimate sales for a new product.
Monthly at a minimum, with actual sales from the closed period. With frequent promotions or short Lead Times, weekly. A model that is never retrained loses accuracy cycle after cycle.
It depends on the category and the aggregation level. WMAPE is the reference metric because it weights by volume: the same percentage error matters more on an A item than on a C item. What decides whether it is acceptable is the cost of the stockout versus the cost of the overstock.
Felipe has specialized in the application of artificial intelligence to optimize supply chains, helping companies to predict demand, manage inventories and determine the ideal times to buy raw materials.
.jpg)
