Flower Inventory Forecast
AI & Data
Semester programme:Artificial Intelligence
Client company:GreenZone
Dai Song
Noah Janssens
Eralp Kurdas
Project description
This project addresses the challenge of improving Greenzone's inventory management by predicting flower demand more accurately. The research question was: How can an AI-driven application support better purchasing decisions by forecasting sales trends and seasonal patterns? Since direct integration with Greenzone’s existing Qlik Sense environment was not feasible, we developed a separate forecasting app. This app uses historical sales data to generate future demand predictions and visualizes them in an interactive and accessible interface. The goal is to help employees, regardless of their experience to also anticipate flower sales, reduce waste, and improve ordering efficiency throughout the year.
Context
Greenzone operates in the flower distribution industry and serves over 1,000 florists across Germany and neighboring countries. Due to the perishable nature of flowers and the highly seasonal nature of sales, accurate inventory planning is critical. Greenzone has relied on employee experience and intuition to determine what and how much to purchase, but this approach becomes unreliable during peak seasons or when onboarding new staff. The company uses Qlik Sense for business intelligence, but technical constraints prevented full integration of predictive models into their system. So as a result, we created an external forecasting app that uses AI (SARIMAX) to identify trends, support purchase planning, and visualize future demand. This standalone tool acts as a decision-support system to increase consistency and operational efficiency, making Greenzone less dependent on subjective judgment and more responsive to market dynamics.
Results
The main result of the project is a working forecasting app that helps Greenzone predict flower demand more accurately. It uses a SARIMAX model to capture seasonal patterns and presents predictions in a simple, interactive interface. Although the original plan was to integrate the model into Greenzone’s Qlik Sense environment, this wasn’t possible due to restrictions on uploading custom Python code or external models within their system—not because of technical limitations on our side.
As a result, we developed a standalone app that runs independently but still supports key business decisions, such as weekly purchasing. The app reduces waste, supports less experienced staff, and brings more consistency to stock planning. It was validated through internal testing, historical sales comparison, and positive feedback from Greenzone’s team. The tool currently fits TRL 5–6: it works in a relevant setting and is ready for practical use but full operational deployment would require further infrastructure adjustments.
Download Business Proposal (PDF)
About the project group
All the team members came from a business background and we did a semester in AI-Core. We held two scheduled meetings per week to recap the progress and plan our next steps. Throughout the project, we had contact with our stakeholder to keep them informed, and the company who was responsible for the data delivery of our stakeholder. Our group collaborated effectively by maintaining a productive working dynamic.