Exploring the technology of machine learning to improve the demand forecasting

Sammanfattning: Title: Exploring the technology of machine learning to improve the demand forecasting Authors: Viktoria Gerdtham & Karolina Nilsson Background: The technology of artificial intelligence is considered to be one of the most important technological advances of our era and a fundamental driver of economic growth in our society. Axis Communications, a customer-oriented technology company delivering end-to-end solutions that strives to maximize growth is dependent on a scalable and flexible supply chain. The technology of machine learning can be beneficial to investigate when considering strategies of improving the demand forecast. Problem description: The motivation of this study derives from Axis’ supply chain setup. It takes approximately four months to procure 85 % of the cameras’ components. Despite this long time period for procurement, Axis offers a lead time of ten days towards their customers. With this setup, it is crucial to have an accurate demand forecast and the current processes facilitating the demand forecasting today are time-consuming, subjective, complex and leads to patterns and trends going undiscovered. Purpose: The purpose of this research is to explore the phenomenon; ML applications in the demand forecasting at Axis. This purpose is fulfilled by answering the following research questions. ● What kind of ML application could improve the demand forecast at Axis? ● What factors should be considered when implementing machine learning into the demand planning processes at Axis? Methodology: A single case study was performed at Axis Communications to research the phenomenon of improving the demand forecasting with machine learning. The study was qualitative and explorative by nature and a research process framework was followed. Furthermore, the study’s trustworthiness was thoroughly analyzed in terms of validity, reliability as well as generalizability. Conclusions: The overall conclusions of the research were three independent model propositions. The models presented in the research are an attribute-based model, a prediction model based on distributor data as well as a parameter segmentation model. The proposed models are believed to generate more accurate forecasts, less manual work as well as more quantitative data processing. Recommendations: The concluding recommendations of the research were the following; • Form a strategic stance determining the level of acquisition in competence and knowledge. • Top management needs to communicate the recognized success stories of initiated projects to counteract cultural resistance and unwillingness to allocate resources. • Clear roles and responsibilities should be set as well as the implementation of data-quality principles, management principles and customized policies within the entirety of the organization to further facilitate the implementation of ML. • Axis’ demand planning department should strive to develop its internal ML competence and knowledge. Knowledge on the most relevant and basic ML would facilitate idea generation as well as generate awareness of benefits and risks. • Axis is suggested to recruit ML competence that can support implementation projects within business processes at the operations department. This recommendation implies competence allocated to support ML implementation into the demand planning processes. • When acquiring ML competence, it is recommended that Axis defines and communicates their level of ambition and expectations to avoid the ML competence leaving due to disappointments connected to the working assignments. • Axis could suggestively develop a ML support network to provide a support system for the implementation team. • Axis should consider the composing of implementation teams to bridge the gap of knowledge and competence between the demand planning team and the ML competence to ease collaboration, streamline the implementations and trigger the idea generation. • Axis should also raise awareness to the limited experience in the implementation and usage of ML in Axis’ business processes. • It is crucial to develop an understanding of how ML models should be interpreted and their limitations to reduce overreliance. • Axis is recommended to strive for shared incentives between functions involved when initiating ML projects. • Develop a joint and generic data warehouse to consolidate the data available at Axis. • Take action towards cleaning data and standardizing the usage of information systems Keywords: machine learning, demand forecasting, business processes, artificial intelligence, operations