Utilizing Large Language Modelsto draw better conclusions fromdata intensive reports: A case study between Impact Solutions Scandinavia andRed Bull
2025 (English)Independent thesis Basic level (degree of Bachelor), 10 credits / 15 HE credits
Student thesis
Abstract [en]
In modern data-driven environments, companies that manage vending machines must processlarge volumes of sales data to gain insights into product performance and customer behavior.This report describes the development of a reporting system designed for Impact Solution, a com-pany that supplies vending machines in partnership with both refillers and well-known beverage providers including Red Bull and Coca Cola. The goal was to build a scalable and flexible reportingframework that could automatically generate tailored performance reports for different providers, based on their individual needs and product portfolios.
The solution combines traditional data analysis techniques in Python with advanced tools such aslarge language models from OpenAI. By using structured data pipelines together with visualizationlibraries like Pandas and Seaborn, the system can present clear trends and market patterns. Anadditional component called Vilmas tankar was developed to automatically generate written sum-maries that explain the results in natural language. These summaries provide insight into marketshare changes, product popularity, and year-over-year comparisons without the need for manualinterpretation.
The findings suggest that a well-structured analytics system, combined with prompt-engineered AIsummaries, can significantly reduce the time and effort required to produce high-quality reports.The approach offers both practical value for everyday business decisions and a foundation forfuture tools that allow users to interact directly with their data in more intuitive ways. The projectalso highlights the importance of aligning technical solutions with the strategic goals of differentstakeholders, to ensure that the insights delivered are both relevant and actionable.
Place, publisher, year, edition, pages
2025. , p. 29
Series
MATVET-F
Keywords [en]
LLM, Data analytics, Vending machines, AI - analysis
National Category
Computer and Information Sciences
Identifiers
URN: urn:nbn:se:uu:diva-560760OAI: oai:DiVA.org:uu-560760DiVA, id: diva2:1972419
External cooperation
Impact Solution
Educational program
Master Programme in Engineering Physics
Presentation
2025-06-03, 4007, Ångströmlaboratoriet, Uppsala, 14:45 (Swedish)
Supervisors
Examiners
2025-08-082025-06-182025-08-08Bibliographically approved