Prediction of consumer behaviour in emergency scenarios (the case of COVID-19)

by | Nov 16, 2020 | 0 comments

Application Track:

Theme Driven




Proposed by:


Entity Logo:

Summary of the entity:

Instituto Tecnológico de Informática (ITI) is a Research & Technology centre specialized in ICT, located in Valencia, Spain. ITI research and innovation activity is developed by a team of 200 highly-skilled researchers and technicians, focused on 6 key digital enablers around the Data Cycle: Cyber connectivity, Cyber-Physical Systems, Computing Infrastructures & Platforms, Big Data Analytics, Optimization Technologies, and Artificial Intelligence. All this knowledge and experience is brought to the industrial and public sector through a range of services: Access to infrastructure and technology platforms, Access to specialist expertise on digitisation & applications, Collaborative research for Industry needs, Demonstration of best practices, Training and Education, Showcase technologies in pilot factories, and Support experimentation in real-life environments.

ITI provides a Big Data Space offering infrastructure, tools and data for research and experimentation with Big Data Technologies, which has been awarded with the BDVA i-Space Silver label and is a key infrastructure for the Data Cycle Hub. The Data Cycle Hub (DCH, is the reference one-stop shop DIH in the Valencia region to foster data driven and artificial intelligence-based innovation. The objective is to bridge the gap between research and industry, specifically SMEs, providing innovative solutions and services that require advanced data analytics, automatic learning and artificial intelligence. The DCH is one of the 3 Spanish members of the European Network of Artificial Intelligence DIHs (, and was one of the 5 finalists to the DIHNET DIH Champions Challenge 2019.

Summary of the challenge:

The challenge is to develop predictive models able to anticipate the key product demands and its peak times during the different stages of a crisis / emergency scenario, depending on customer profiles.


Stakeholders: Customers, Retailers


In times of emergency scenarios or crisis such as the current pandemic, data science can be really helpful for predicting consumer behaviour, monitor changes, and anticipate to them to optimize the supply chain, stocks, and reduce waste. When people go through an economic crisis or disastrous scenarios like COVID19, they change their buying patterns depending on the evolution of the situation. Previous studies have identified some behaviour indicators about how grocery shopping changes in past crisis: for instance, items for quick meal preparation were the most bought during the most difficult times, but decreased during recoveries. Identifying which are the key items that retailers must provision when an emergency scenario arises is key, not only to optimize supply chains and logistics, but also to avoid panic buying when item shortages are perceived.

The challenge is to develop predictive models that could identify the buying patterns of consumers during a crisis/emergency scenario, and identify which products need to be prioritised and need further stocks. To this end, consumer behaviour as well as relevant data needs to be processed and analysed to identify relevant item categories, significant customer profiles and buying patterns, and the evolution of the buying patterns (e.g. item preferences per customer profile) along the different stages of the crisis (e.g. early crisis, peak of the crisis, recovery, back to normality). Relevant data that could be considered are: buying transaction data, consumer profiles, consumption statistics, buying trends, social media, etc. For retailers, these predictive models will be a key forecasting tool for improving costs planning (e.g. rental and investments management, inventories optimization) as well as supply chain optimization (e.g. supply chain providers selection, logistics costs, etc.). The goal will be to anticipate the peak times and key product demands, according to the socio-demographic characteristics and expected mobility surrounding the physical stores.


  • REACH Data Providers
    • MIGROS: Sales data. Provides aggregated sales transaction data from MIGROS stores in Turkey, including fresh produces (store, product, sales amount, sales quantity, date).
    • SONAE: Customer transactions, Customer DNA. Two datasets from SONAEMC stores in Portugal, that provide sales transaction data and relevant segmentations on customer level for the identification of different profiles.
  • Other datasets: Other datasets could be used if needed to improve the building of the predictive model. For instance:
    • Statistics on consumption demand, at National or Global level, focusing on crisis or emergency periods, like COVID19 or previous ones.
    • Social media (identification of major buying keywords)
    • Google Mobility index (correlation with sales)
    • Historical Weather conditions (correlation with seasonal trends).

Expected outcomes:

  • Better knowledge of COVID19 impact in consumer behaviour within the retail sector
  • List of relevant item categories and consumption patterns during the evolution of an emergency scenario
  • List of customer profiles and evolution of its buying patterns over time within an emergency scenario
  • Predictive model or tool for supporting marketing and sales in demand forecasting when facing uncertain scenarios

    How do we apply?

    Read the Guidelines for Applicants

    Doubts or questions? Read more about REACH on the About Us page,

    have a look at our FAQ section or drop us an email at