Predictive Inventory Management
for Fresh Dairy Products in Retail

We are pleased to announce a Data Challenge dedicated to the development of predictive models for inventory management in the retail sector. The initiative aims to foster applied statistical research and collaboration between academia and industry, promoting data-driven solutions for more sustainable retail supply chains. The challenge focuses on fresh dairy products—such as milk, yogurt, and fresh cheeses—sold in the large-scale retail sector (GDO). This product category presents several operational and analytical challenges:

• High product turnover and short shelf life (typically 7–15 days)
• Strong demand variability, influenced by day-of-week patterns, seasonality, and promotional campaigns
• Logistic dependence on external distribution centers
• Significant waste and product returns, particularly during periods of low demand or excessive promotions.

These characteristics make fresh dairy products an ideal testbed for statistical and machine learning approaches to predictive inventory management.

Dataset

Participants will work with a real-world dataset provided by Astrea Consulting Srl, derived from a network of retail stores operating within the Conad and Todis distribution channels. The dataset includes information on:

• product identifier,
• store identifier,
• inventory movement date,

for 30 products, each observed over the January 2023-December 2025 time span.

For each inventory movement, the following variables are available:

• initial stock quantity,
• purchased quantity,
• sold quantity,
• final stock quantity,
• quantity sold under promotion,
• presence of sell-in / sell-out promotional campaigns.

We gratefully acknowledge Astrea Consulting Srl for making this dataset available and for supporting this initiative that bridges academic research and real-world retail analytics.

Challenge Objective

The goal of the challenge is to develop unsupervised predictive models capable of identifying inventory conditions for each product in each store. Based on historical inventory movements, participants must classify each observation into one of three states. The predicted classifications should be provided in a .csv file containing a single column with 229,499 rows (one for each observation), without a header. The three categories must be encoded as integers:

• In Stock,
• Out of Stock,
• Over Stock.

Eligibility and Team Composition

• Teams may consist of up to three members,
• All participants must be 30 years old or younger at the time of submission.

Timeline

• Dataset release: 1 April 2026,
• Submission deadline: 8 May 2026,
• List of the five selected teams: 15 June 2026,
• Workshop registration: The team leaders of the selected teams are requested to register by 30 June 2026 and attend the workshop,
• Presentation of shortlisted projects: 27 August 2026 (second day of the MBC2 Workshop in Catania).

Evaluation Process

To ensure both methodological rigor and practical relevance, all submissions will be evaluated by a blended academia–industry committee, that will select the short list of five best projects according to the following combined criterion:

The leaders of the five best projects will be invited to present their work during a dedicated session at the workshop. The final ranking will be determined by a majority vote of the workshop attendees. The winning team will be announced during the workshop dinner (27 August, h. 20).

Awards

• The winner  team will receive a  prize of € 1000.
• The second-place team  will receive a  prize of € 700.
• All the five team leaders are offered the workshop dinner (27 August, 2026).

Registration

Teams must be preliminarily register for the Data Challenge through the Registration webform. :

Registration must include:

• the composition of the team and the team name.
• the Non-Disclosure Agreement signed by each component of the team (form available here).

During the registration process, participants will be required to upload a PDF copy of the payment receipt/proof of payment. A zipped file containing the technical report, the vector of predicted classification and the software script must be submitted through the Submission webform.

Data challenge registration fee

Registration fee for the team to the data challenge: € 60. The payment of the fee must be made by either:

• Bank transfer:
  Bank name: ALLIANZ BANK FINANCIAL ADVISORS SPA
  Address: Piazza Tre Torri, 3 - 20145 Milano (Italy)
  Account holder: COMITATO ORGANIZZATORE MBC2
  BIC/SWIFT code: BKRAITMMXXX
  IBAN code: IT 71 H 03589 01600 010570886048
  Bank account: 010570886048
  Bank transfer reference: MBC2 2026 Registration fee – Last Name First Name
• Paypal
  .

Data availability

Data will be available to candidates  after the registration.

Last update:  31 March 2026.