Forecasting with machine learning – Have Finnish controllers adopted advanced technology in sales forecasting?

Loading...
Thumbnail Image

URL

Journal Title

Journal ISSN

Volume Title

School of Business | Master's thesis

Date

2021

Major/Subject

Mcode

Degree programme

Accounting

Language

en

Pages

84

Series

Abstract

Technology has enabled usage of different advanced technological applications and has shaped and will continue to shape our way to work (Acemogly & Restrepo, 2019). Machine learning which lies in the heart of artificial intelligence is “one of today’s most rapidly growing technical fields…” (Jordan, 2015). Predicting future sales plays a key role in maintaining a profitable and growing business. For years, organizations have attempted to improve their sales forecasting by taking advantage of computer and information system technologies (Moon et al., 2003) and controllers are expected to have a sufficient data analytics skill set (Oesterreich & Teuteberg, 2019). This leads to the hypothesis that controllers are using advanced tools to make better insights about data. Academic literature about machine learning is still quite young but growing fast. Accounting literature emphasizes benefits and challenges what advanced technology brings and is also future-oriented. Extensive practical studies are still under development as many studies focuses on individual case studies. As this study focuses on machine learning in forecasting it also tries to expand the perspective and study how controllers are using various advanced tools in different Finnish organizations. This approach sheds light to controllers forecasting practices and the current state of advanced technology among Finnish organizations. Against the expectation of hypothesis is seems that controllers have not implemented advanced tools in practice. Organizations are still taking the first steps with these methods and are optimistic that the adaptation of more advanced technology will increase in the future. However, it seems that there are some enormous obstacles which organizations are required to solve. Especially, organizations are lacking skilled people, are operating with old-fashioned systems which do not support more advanced tools, and they do not have the required IT infrastructure in place. Moreover, successful business cases around machine learning are still missing. The conclusion of the theoretical part aligns well with the results of the empirical study. Organizations are still trying to figure out the best tools to be implemented for forecasting. It seems that simple models such as regression models have been outperforming the advanced models during the past few decades. This mirrors the overall atmosphere in practice: why make large investments to advanced technology when current methods provide “good enough” results. Qualitative forecasting technique is still the common method. Combining it with different statistical methods are considered to make better results than machine learning, at least in a volatile market environment. All in all, there is a lot of progress to be made for machine learning to gain its superiority.

Description

Thesis advisor

Sihvonen, Jukka

Keywords

controller, forecasting, machine learning, artificial intelligence

Other note

Citation