Retail has been one of the industries most impacted by the advancements of artificial intelligence (AI). Shopping is becoming more and more frictionless to the point where customers can simply say “order more garbage bags” and their home virtual assistant can do the rest. Product warehouses are highly automated and soon regular everyday purchases will be delivered by an autonomous car or drone. Shoppers can now enter a brick and mortar store, or shopping area, scan a credit card or loyalty card, take what they want off the shelves and walk out without having to checkout at a counter. All of these new advancements are made possible by artificial intelligence.
Retailers are using AI to better forecast product demand
, guaranteeing that stores stay in stock with products their customers want. Meanwhile, machine and deep learning models can be used for marketing purposes, providing fine grain customer segmentations
that can be used for next best offers
and personalization. Lastly, online retailers are providing tools for their customers to search for products with pictures of similar products or companion products. Some retailers even allow customers to upload picture of themselves in order to get a recommendation for items that look good with the outfit they have on.
Outside of the retail industry, one of the most widely used applications of artificial intelligence is detecting and tracking objects in video images. For retailers, this provides an opportunity to combine data from tracking people, using their in-store cameras, with their operational data - like sales data, store layouts, customer data and inventory-to dramatically increase sales and lower costs. We refer to analytics that combine operational data with video metadata as video analytics.
Through the work Teradata has done with retail customers, we have developed a number of video analytic capabilities that can help retailers solve some of their biggest challenges.
- Path Analysis tracks customers or sales associates as they move around the store with much more accuracy than sensors or beacons. By combining the path customers take through the store with POS, retailers are able to understand the path customers take leading up to a purchase. Retailers can also use path analysis to plan where in the store to place products or signage for customers to engage with. One of the challenges with path analysis is being able to calibrate multiple in-store cameras and to track customers when they become blocked from the cameras’ views.
- Dwell Analytics monitors the amount of time customers spend in a particular location, which may indicate inefficiencies in the process of servicing customers, or that customers are not able to find what they are looking for.
- Queue Analytics monitors the amount of time customers spend in a line waiting to make a purchase or have someone help them.
- Sentiment Analysis uses facial recognition to determine a customer’s sentiment, such as when they become frustrated waiting in line to make a purchase or when they appear delighted with their in-store experience.
- In-Store Engagement detects when a customer is engaged with a product, signage or sales associate. This could be determined by analyzing the amount of time spent in a specific area, or their proximity to another person, or detecting when a customer touches or picks up a product.
- Associate Time Motion tracks and analyzes sales associates’ activities, such as how much time they spend engaged with customers, how much time they spend on the store floor vs the back room, how much time they generally spend performing value added activities. One of the key challenges is to be able to differentiate between sales associates and customers.
By combining these video analytic capabilities with other data and analytics, retailers can optimize their store layouts, increase sales with higher conversions, better forecast the number of associates they need, and optimize their in-store processes to lower costs and improve customer experience. As an example, think of a retailer with thousands of stores, they want to determine the optimum number of sales associates in 30 minute intervals across all of their stores. The retailer suspects that they are currently overstaffed, but they do not want to increase the amount of time customers spend waiting for an associate to help them, which would lead to lower conversion rates. With over $100M in annual staffing related costs, even a 1% decrease in staffing levels could save $1M a year.
Another application of video analytics is to do in-store A/B testing. Retailers want to know - just as they would on the digital channel - what store layouts work the best and what is the best place for signage or product placements. Currently, the time and cost to test different layouts is prohibitive. With video analytics, retailers are able to quickly and accurately measure alternative layouts on demand. By testing and identifying the best placements for products retailers can increase sales and improve the customer experience.
provides a key component in enabling these capabilities. We combine advanced analytic approaches like Convolutional Neural Networks
(CNN), Single Shot Detectors
(SSD) and Kalman Filters
, with Vantage’s 4D analytics
to calculate different metrics, like the amount of time a customer spent in a certain location or the path taken in a store over a selected period of time. Finally, by combining this with operational data, such as whether or not a customer actually bought something, businesses are able to get answers to questions that impact their bottom line.
The examples above are only a few ways video analytics can be applied to retail stores. Video analytics is another step in the incredible journey the retail industry is experiencing as a result of artificial intelligence. For retail businesses, there is an amazing opportunity with Vantage to unlock the value of their video data by combining it with operational data, enabling them to deliver transformative business outcomes and answer the most difficult strategic questions. Are you ready to use vision analytics to give your business a competitive edge?
Peter is the AI Team Lead for Americas. He is responsible for the successful delivery of AI projects in America, supporting the sales field and managing the team’s program of research and IP development. Previously, Peter was the director of services at Think Big Analytics for 5 years, responsible for the delivery of big data and advance analytic projects. Peter has a strong background in program management and has successfully delivered large programs of work in a range of different industries. Peter holds a bachelors of commerce in management science and a master’s in computer science from McGill University.
View all posts by Peter Mackenzie