Modelling of shoppers’ attention patterns and behaviour in convenience stores
We wish to model the behaviour of shoppers searching for a specific product (or set of products), as indicated by which product(s) they purchase. We propose to use shopper body pose/skeleton information to identify specific activities, such as turning to view an item on a shelf, reaching for items, putting objects back, moving around obstacles, exploratory movements vs. direct homing in on product locations. We will use eye-tracking, in a controlled study, to determine what places in the store and which objects people look at when searching for specific products. We will model and detect cases of distraction, where shoppers are affected by (viewing or reaching for) items that they do not buy. We will analyze behaviour of shoppers doing foraging, which is the collection of multiple articles. Do they optimize the path between purchased articles, or do they do independent search patterns for each individual product? We will also investigate whether shoppers’ attention patterns can be clustered into groups, allowing personalized models to be created.
Graduate students from the Department of Electrical and Computer Engineering
Yinan Wang - PhD
Farzaneh Askari - PhD
Rezvan Sherkati - PhD
Xiangyu Li - MSc
Sansitha Panchadsaram - MSc
Principal Investigator: James Clark
3D Digital Twin
During the pandemic, it has become clear that online shopping is becoming crucial to retail. For the RIL, we plan on taking this idea one step further. Our goal is to develop an immersive virtual reality (VR) simulation of the RIL, as well as a tool to aid physical shoppers with additional information on their mobile devices (augmented reality, AR). Through the VR version of the RIL:
- we can show customers additional information for the products they are interested in (highlight deals, nutritional info, related products);
- we can play back anonymized recordings of past customers to inspect their behaviour and train an algorithm (reinforcement learning agent) to behave like a customer;
- with this, we can test different store layouts and product configurations in the store; and
- ultimately show these to VR shoppers. The findings can then be transferred back to the physical store;
- we also have the opportunity through the VR simulation to try different ways of incentivizing purchasing behaviors, e.g., for people who are looking to buy more sustainably or healthy.
Postdoctoral Researcher: Florian Golemo
Principal Investigator: Derek Nowrouzezahrai and Maxime Cohen
Evaluating the unique experiences and decision-making processes of people with type 2 diabetes when grocery shopping
The purpose of this study is to evaluate the experiences, decision-making processes, and needs of people living with type 2 diabetes (PwT2D) when shopping in a retail food outlet, compared to consumers without diabetes. While diet is a primary risk factor for T2D, complex societal factors, such as the retail food environment, are important contributors to food choices and eating behaviour. However, little is known about the impact of retailing factors on the food choices of PwT2D specifically. Being diagnosed with T2D can have a profound impact on the food choices of patients, as they must take into account their disease management and nutrition-related knowledge.
PhD student from the Faculty of Human Nutrition
Pablo Arrona Cardoza
Principal Investigator: Daiva Nielsen
Robotics for retail
The nature of this project is exploratory and aim to find potential synergy between human-robot interaction and retail innovation through series of brainstorming sessions, trial, and error processes. As such, the final number and type of studies (e.g., controlled human-subjects’ experiments, technical developments) involved in the project will be defined over time. The guiding principle of this work will be to leverage the newly acquired interactive robotic platforms at the RAISE Lab and the unique living lab environment present at the Retail Innovation Lab.
Principal Investigator: AJung Moon
Edge-A-Eye project
The specific focus of the McGill team within this project is to develop technology that supports acquisition and verification of a desired object once the user is situated in a limited radius of the target item, so that it can be seen by a smartphone camera from that position. This will involve dynamically directing the user to reach the object, verifying that the intended item has been acquired, and providing desired product information such as ingredient listings or price. The technology is intended to run on commodity smartphones, potentially in conjunction with bone conduction headphones for auditory display. As a core focus of the McGill team’s activities aiming to overcome some of the barriers facing those with BVI in the retail environment, we will be tackling the sequence of “scan, locate, interact, verify” tasks, predicated on the use of a commodity smartphone, leveraging the power of edge computing and artificial intelligence models.
Principal Investigator: Jeremy Cooperstock
Incentivizing healthy food choices using add-on bundling: A field experiment: Part 2
How can retailers make promotional decisions for healthy vs unhealthy products? Finding the right promotion for the product is very important for retailers to maximize the sales and profits. This field experiment investigates add-on bundling and discounting strategies across two product categories: healthy and unhealthy. Our study encompassed six types of promotions:
- Unhealthy bundle (coffee purchase with the option to add a pastry for $1.50),
- Healthy bundle (including a healthy snack as an add-on for $2.50),
- Unhealthy discount (any pastry for $1.50),
- Healthy discount (any healthy snack for $2.50),
- Choice bundle (offering either a pastry for $1.50 or a healthy snack for $2.50 as an add-on), and
- Choice discount (any pastry for $1.50 or any healthy snack for $2.50).
Through this comprehensive approach, we aim to unravel the nuanced dynamics of how these promotions influence consumers' decisions, shedding light on effective strategies for retailers to promote the products in the best way.
PhD student from Desautels Faculty of Management
Nymisha Bandi
Student co-supervisors: Saibal Ray and Maxime Cohen
Reinforcement Learning for Retail Counterfactuals
IVADO-funded project
The motivation of this project is to use reinforcement learning on the retail innovation lab in order to develop a virtual store environment that will allow us to train virtual agents according to real-world customer trajectory and purchasing behaviours. These virtual agents will allow us to perform customer-and store-centric configuration and layout experiments at scale, before deploying them in the physical store. An example of one such experiment would be the optimization of store layout that promotes healthy consumer buying behaviors.
