Artificial Intelligence in Automatic Retail Store

The retail industry has realized a significant increase in the number of customers while computers are getting stronger and better trained to analyze age differences in disparate regions all over the world.

Basing on the data collected from customer, retailers can find out appropriate business strategies to increase sales.

Retailers have been testing activity detection to get an insight in customers’ behaviors in the store. For instance, activity detection can be applied to understand customers and distribute product types that suit specific customers as well as deliver strategic discounts.

Other AI use cases in retail include better predictive and prescriptive models — essentially answering forward-facing questions like what will possibly happen in a particular store tomorrow or this weekend — and automatic recommendations — along the lines of where new stores should be opened, what the layouts should entail, and what products should be stocked in the store and where. Shoppers’ needs and values will be better met when retail combines art and science.

What are key features for future Smart Retail?

In partnership with Intel, a technology enterprise of Vietnam has applied Intel’s technologies in conjunction with our key technologies for 3D image processing to address Intel’s Store retail problem.

According to Intel, a Smart Retail should have the following functions:

Ad Metrics / Dwell Analysis – advertisement performance metrics and dwell time analysis focus on customer interactions with products and promotions.

Face Recognition – with the precision of Smart Retail’s face recognition – be alerted to known shoplifters or VIPs as they enter a retail location with Smart Alerts.

Customer Count – knowing how many customers enter in and out of stores is valuable information and can be used to put other measures, such as gross sales, into perspective.

Customer Metrics – customer metrics captures what credit and membership card swipes miss; all the customers that did not purchase anything and more quantified, qualitative data on those that did.

Intrusion Detection –monitor secure areas of your back office or warehouse with unauthorized entry detection or loitering.

Queue Manager – manage your retail floor with in-depth analytics that monitor POS stations and other areas. This data can trigger alerts that notify internal teams of longer than usual wait times to allow better service of your customers and reduce unnecessary lines.

Store Conversion – store conversion gives context to sales numbers. Smart Retail’s store conversion feature can accurately keep track of those who enter and purchase and those who do not.

Traffic Map – see how consumers move about the premises. The system tracks where people are spending time and which products or services they come in contact with

Solution for S. store Intel’s problem

S. retail stores are aspiring to be the smartest retail store in Asia and are looking into the below initiatives to ensure that:

S. is able to identify the customer before or during the customer’s journey to the shop t

Issue electronic queue ticket or allow customer to book an appointment to store
before the visit

Identify the customer in the store and monitoring the customer’s activities and
intelligently push personalized offers/promotions to the customer

Ability to detect the customer’s facial expression and solution a notification method to address customer’s unhappiness

Smart analysis of the profile of the customer base who visited the store for retail
planning on targeted promotions based on the customer profile

Intelligently locate the customer when it is the customer’s turn to be served

Technology Selection

To achieve highest quality for the system, we recommends Intel’s 3D cameras such as R200 for customers tracking. For IoT data processing, we recommends AWS platform for data ingestion and processing using AWS IoT service, AWS Elastic MapReduce (EMR) because AWS platform offers a highly available, scalable and low cost services.

The system consists of 4 main components:

Smart Appointment System

Customer open MyS. App to book an appointment,

Once the customer near a shop and within the appointment date, the system will send a notification to the customer to get a Ticket Queue,

The customer open the notification, the system will create an electronic ticket for the customer.

In-store Camera System to identify customers and track Customer’s location

Customer walks into a S. store,

The camera system capture the customer’s 3D images,

The camera system check the customer’s identity from S. CRM image database,

The camera sends customer’s location data along with timestamp to S3 on AWS Cloud in Singapore

Personalized Recommendation System

The recommendation system ingest customer’s data in real time as the customer’s data comes into the system via Kinesis,

The recommendation system analyzes and pushes personalized promotion to customer’s MyS. App according to rule engines specified by S. business users.

Customer receives notifications and open MyS. App to see promotions.

MyS. App can shows 3D map of the store and highlights promoted items on the map.

Crowd Analytics System (CAS)

S. business user (SBU) specify the store ID, customer ID, then click on Query,

The CAS show how long the customer stays in the Store, filtered by date range, time and by area,

SBU select a store ID, then click show movement statistics, the CAS show how many customers move from area 1 to area 2,

SBU select a store ID and an area, then the CAS system can show how long customers stay in that area on average, maximum or minimum.

The system will be designed on top of the followings:

Supporting multiple user platforms.

A central API to handle all back-end processes. The API must be stateless and scalable.

Using reliable, off-the-shelf 3D cameras to handle the job of capturing 3D data of customer’s movement.

Indoor customers tracking will be done by fusing 3D data from multiple cameras to achieve highest accuracy level of tracking without being too intrusive such as explicitly asking S.’s customers to open a phone.

dx future


Leave a Reply