1,800+

Parkland gas stations

25

countries

22+ billion

liters of Parkland petroleum products marketed

clients_ThermoFisherLogo

EXECUTIVE SUMMARY

TensorIoT developed an edge enabled, machine learning model, capable of monitoring the arrival and departure of cars from defined areas of interest (pumps and parking spots), providing traffic metrics as an output on a custom built user interface.

GOAL

How can you use existing camera systems to recognize, record, and track customer traffic to optimize store inventory, layout, and signage decisions?

RESULTS

The solution proved capable at monitoring vehicle traffic in the forecourt, counting the number of customers entering/exiting the store, and tracking in-store foot traffic on a heatmap to help optimize store inventory, layout, and signage decisions.

Background

Parkland Fuel has begun the process of leveraging cutting-edge digital technology to better serve, understand, and influence customer behavior in and around their convenience stores. Building on the success of the initial PoC projects, Parkland now wishes to build out additional use cases that will start providing broader business impact. Leveraging the existing hardware from the previously developed solution, new computer vision models will be trained and deployed to Parkland’s Panorama device as part of an edge application, complete with requisite business logic providing Parkland with additional data about their customers, their experiences, and how best to improve them.

The Challenge

The client had no data or system in place to support regular business insights and planning such as average out of stock rates, inventory demand forecasting, etc. TensorIoT needed to implement a computer vision system to track inventory in its stores, what are trends over time, what is in/out of stock, where products are being sold in the store, impact of seasonality, and so on in their retail stores.

The Solution

TensorIoT built a data-backed understanding of the customer traffic in and around their hundreds of gas station convenience stores leveraging computer vision with AWS Panorama. This proved capable at monitoring vehicle traffic in the forecourt, counting the number of customers entering/exiting the store, and tracking in-store foot traffic on a heatmap to help optimize store inventory, layout, and signage decisions. Parkland now has the ability to extract customer demographic information and perform customer sentiment analysis from store video. Further, they can analyze both employee utilization to indemnify reoccurring idle periods and queue length to identify problematic periods.