I am often asked whether Artificial Intelligence can help to achieve better sales forecasts and procurement planning. My answer: Yes, it certainly can, and especially the improvements by using machine learning algorithms can be really impressive. To achieve this, two important things must be taken into account from the very beginning of a project:
- you must develop a good feeling for the “external” driving factors that significantly influence your demand planning,
- your logistics chain must be able to respond to the improved forecasts of future sales.
What exactly does this mean? Let’s have a detailed look at these two points.
More Barbecue in Summer
External factors include information that is digitally available and influence your demand. AI can recognize these correlations between data series and learn how to assess them.
An easy example of such an external factor is the weather forecast. Good weather on a summer weekend can improve sales of grilled meat, charcoal and ice cream. Especially fresh products in the food industry, which are always noticeably influenced by the weather. AI can help to improve the quality of forecasts by including this data into a demand planning process. In our example, forecasts can become 50% more reliable than with more classical methods of pattern recognition used for the analysis of the outgoing stock data. This means that the deviation of the actual sales and the forecast can be halved again by using Machine Learning. This is an important aspect, especially in the discussion of sustainability and in margin-limited businesses! Cannibalism effects (i.e. the influence of similar articles on each other) can also be learned by an AI. For example, by learning the effect of actions on an article (organic broccoli) for the compound of similar articles (other broccoli varieties) on the basis of historical data.
But pay attention! When searching for these external factors, less is more. To be honest, no one can expect AI that just has internet access to learn independently, as factors from the Big Data world influence the demand in a certain way. Instead, it is a great advantage to have fixed assessment of the main factors and have them digitally available. That’s why human contribution is important during an AI project. In our example, weather data will probably have little influence on the slow-moving assortment of a DIY store. In the garden plant department, however, the situation may be different. Finding the right factors and having them available in digital form to feed the existing AI technology is key for a successful machine learning project.
How agile is your logistics?
The second point is as important as it is challenging for your business. It is about how a company can react to a forecast that has been changed by AI. Sticking with our earlier weather example: The weather forecast today is only really good for a maximum of three days in the future. After that it is more a rough guideline than a reliable forecast. This means high demands on your logistics and supply chain, as it must cope with possible changes in demand planning within this short time frame. Food retail chains are usually fast enough, as they have to plan for perishable food on a daily basis anyway. They can usually adjust their assortments in the stores within 24 hours. However, not all companies can or must react so quickly. Replenishment times in the automotive industry or technical wholesale are often long due to global procurement channels. At the same time, products and goods do not have a minimum shelf life and are often stocked for a long time. Apart from the fact that inventory management or even the calculation of safety stocks should never be carried out without an intelligent planning system, AI algorithms can improve forecasts and provide a more reliable view of the future, if external factors also provide information about long-term demand developments.
To be precise: The more accurate and stable the data of the external factors are for the distant future, the better they can be used for more time-consuming logistics chains. The forecast could also be significantly improved over a longer planning horizon, which would have a positive effect on inventory management and warehousing. Yet, without the ability to react to the temporal impact of external factors within the logistics chain, an AI project will not bring the desired success.
Closing Thoughts
When you start an AI project in sales or procurement planning, you usually need to clarify the two factors mentioned in this article. If the associated questions can be answered positively, the chances of an attractive success are high. However, there are many other questions to be answered during the course of an AI project. For example, which AI algorithm is used. Neural networks are good, but not always the best choice for all fields of application. There is more. Therefore, it is important to have a partner at your side who has the necessary AI know-how and who also contrasts the benefits of machine learning algorithms with the effort for a corresponding project. For many questions regarding the improvement of sales and procurement planning, mature optimization systems are already suitable. Based on mathematical algorithms, they can increase the prediction accuracy many times over and provide the planner with optimized order proposals. Service levels that increase to 99% with simultaneous inventory reduction are not uncommon here. Machine Learning, with its constantly growing access to data, gives us the opportunity to go one step further and make our forecast models even more future-proof through artificial intelligence. I see great potential in this.
How reliable is your sales and procurement planning today? Do you already use specialized software?
For the original article, click here.
Image source: Shutterstock, Thanakorn.P
