Predicting Sales Dashboard
What was the question?
Data can be used as a kind of crystal ball. With the help of data, predictions can be made. Lumission has asked us to forecast their sales for the upcoming year. They can use these predictions to better anticipate future sales and make informed decisions.
Fictitious data is displayed in the demo dashboard.
What was the solution?
DATA KINGDOM has made forecasts for eight different product groups using three different methods. In addition to the models, we also present the evaluation criteria, making the accuracy of the models transparent.
The first model is Exponential Smoothing, a simple yet powerful method for forecasting time series. This method assigns weights to previous observations, with more recent data carrying more weight than older data. Exponential Smoothing has three variants: Single, Double, and Triple Exponential Smoothing. Single Exponential Smoothing does not account for trends or seasons, while Double includes trends, and Triple accounts for both trends and seasonal patterns. This model is especially effective for short- to medium-term forecasts and is flexible in applying trends and seasons depending on the dataset.
AutoRegressive Integrated Moving Average
ARIMA (AutoRegressive Integrated Moving Average) is a popular method for time series forecasting that combines three components: autoregression (AR), integration (I), and moving average (MA). The AR component determines how many past values are used for the prediction, while the I component makes the data stationary by taking differences between data points. The MA component looks at errors in previous predictions to improve future values. ARIMA is powerful because it can model both trends and autocorrelations in data, but it requires the data to be stationary. For time series with seasonal patterns, the modified version, SARIMA, is used, which adds seasonal components to the ARIMA model.
The Croston model is specifically designed for forecasting intermittent demand, meaning demand that occurs irregularly and with low frequency. This model splits the time series into two components: the size of the demand (only non-zero values) and the interval between demand occurrences. Based on these two elements and a smoothing parameter, the forecast is continuously updated with each new non-zero value. While Croston doesn’t account for trends or seasons, it is effective for short-term forecasts and ideal for inventory management of products with irregular demand. The model is less accurate than ARIMA or Exponential Smoothing, but it can handle periods without demand, making it ideal for managing inventory of rarely sold products.
This diversity in models allows Lumission to receive tailored forecasts for different product groups, depending on their sales behavior, while taking trends and seasons into account.
What are we looking at?
Data from:
– Navision
For example:
– Monthly forecast
– Weekly forecast
– Average actual sales
– Sales forecast
– Actual sales
– Confidence interval
What does Lumission gain from this?
Lumission’s Sales Dashboard has provided insight into potential future sales. Seasonal patterns are also made clear through the data. With these new insights, Lumission can proactively make decisions that significantly improve the efficiency of their operations.
By adjusting purchasing and inventory strategies in a timely manner, Lumission can prevent shortages or surpluses, thereby optimizing their cost structure. Additionally, they can respond more effectively to market changes and customer needs, which not only enhances operational efficiency but also contributes to improved customer satisfaction. Ultimately, this allows Lumission to not only save costs but also increase profitability by making more informed, data-driven decisions.
How do we work?
To ensure that your finance dashboard perfectly aligns with your organization’s needs, we follow a carefully structured process:
Data Strategy Session: In this session, we focus on gathering insights. What is the company’s overall strategy? Is there a data strategy in place? What are the key KPIs and their definitions? Which data sources are available? Together, we discuss the design requirements for the dashboard.
Inspiration Session: Using the input from the Data Strategy session, we create an inspiration dashboard based on dummy data. During the inspiration session, we present the dashboard and refine the requirements and definitions further based on feedback from the Data Strategy session.
Implementation: We then proceed with connecting the necessary data sources (source systems, databases, Excel, APIs), modeling, visualizing, setting up governance, and providing training. Most importantly, we work together to ensure that data-driven decision-making is embraced throughout the organization.
