Retail video analytics to encourage stronger customer engagement and drive increase in sales

Overview

The client was a grocery chain who had stores in multiple cities and were well known for their products. They wanted to analyze and plan their store layout that can help their customers pick items quicker and easier. They wanted to know what kind of product placement converts well. Assisting customers at the right time in their store was their priority. They wanted to make sure their support staff are available to help the customers in times of need. They wanted to develop strong customer engagement by planning promotions and discounts based on the purchase times and patterns. They wanted to gain a competitive edge by delivering valuable service to customers and increasing their brand loyalty

Challenge

We had the challenge of analyzing CCTV footage of the retail store to study the shopper behaviours and sentiments. The customers had to be classified according to their gender, age and demography. In order to build a persona of the shopper, we also assessed their visit frequency, buying frequency and their products of interest. Generating relevant customer behaviour information from CCTV footage and analyzing them was the key to building the system. These metrics help estimate the customer footfall, consumer-type estimation, dwell time etc.

Solution

This project involves the integration of various Computer Vision algorithms namely person detection, face emotion recognition, age - gender detection and custom object detection. Of all the above-said algorithms, only the custom object detection of a retail stack of consumables required custom training. For the rest of the algorithms, we integrated pre-built models from OpenVINO toolkit from Intel. The data acquisition phase of the project involved collecting images of the retail stack of consumables under varying lighting conditions and camera noise conditions. This was then followed by an object annotation phase with the CVAT tool to generate a segmentation dataset of 2000 images. We implemented a neural network training process derived from the CNN architectures including the Yolo, Xception and Inception. From our performance accuracy studies, we observed that the Yolo V2 Lite model was performing adequately in terms of both accuracy and execution time. We then developed a CCTV video stream processing application that integrates the output from the various object detection algorithms and implemented a robust reporting visualization program. The video stream processing server was then deployed as an on-premise installation that integrated with their existing CCTV network video recorder system.

Impact

We designed a retail analytics solution for their grocery store operation which gave unique insights about consumer interests, buying patterns and frequencies. We developed this project in a time frame of 30 weeks. With our solutions, our client was able to increase their customer engagement by 30% by strategic store layout and quicker checkout experience. The customer pattern gave insights to timely planning of promotions and discounts and they were able to drive their sales up by 25%. They also found a notable 15% increase in shopping experience as the customers were very pleased with the staff involvement and help. Indium (https://www.indiumsoftware.com/blog/video-analytics-for-retail/) reports that their clients saw a solid 15% increase in customer satisfaction post implementation of their retail analytics solution

Technology stack