October - 2002 - issue > Technology

THE TECHNOLOGICAL ADVANCES IN THE LAST century have created new economic paradigms where businesses big and small compete to gain market share and expand operations. In this dog-eat-dog world, businesses constantly have to innovate and adapt new business processes and tools to stay ahead of the competition and reap profits. With an equal-opportunity global marketplace and eroding customer retention initiatives, both brick-and-mortar and online businesses are seeking solutions to unlock the secrets of customer behavior for converting them into loyal buyers of good and services. Technologies like information system management and CRM (Customer Relationship Management) are enabling retailers like Wal-Mart, Best Buy, and eBay to stay competitive in the digital economy by helping them analyze customer transaction data and discover new business opportunities and threats. Though it is much easier for online retailers like Amazon to obtain customer information through web monitoring tools and CRM, the majority of brick-and-mortar retailers are still lagging behind in adopting new business processes and technologies to understand customer behavior and patterns. Current retail intelligence relies partly on market research firms to determine segment-level behavior and partly on heuristics developed about customer segments over time. These techniques tend to be static, slow, incomplete, and expensive and do not provide the timely insight needed for improved decisions and interactions with customers. Thus, although brick-and-mortar channels account for more than 90% of the retail, they have not reaped the potential benefits of CRM technologies and customer analytics.

Computer Vision helping Retailers

Thanks to recent breakthroughs in “Computer Vision”—technologies that obtain information and intelligence from still images and video sequences—we now have the potential for acquiring automated data on customer behavior and composition. These computer vision technologies can work with in-store security cameras for understanding the behavior of customers as they shop. Combining this with sophisticated real-time facial feature analysis techniques, their demographic composition can also be extracted automatically from the camera images. The statistics that are generated, without invading the privacy of individual shoppers (by not uniquely identifying the individuals), is valuable for determing the store traffic, dynamic customer composition, service efficiency, marketing effectiveness, store layout effectiveness, and so on.

Consider this example: company XYZ has launched a $2 million ad campaign for a product targeted towards a particular consumer demographic segment. Traditional techniques to measure the success of the campaign would involve user studies and focus groups. These techniques are flawed by the composition of the focus groups, time of research statistical sampling errors, and so on. Instead, if we have visual sensors installed at cash registers and other hot spots within stores, we can use computer vision techniques to obtain customer behavior and demographics in real-time from the actual shoppers as opposed to a small focus group. The implications of collaborating with the POS (Point of Sale) data are tremendous. It is now possible to identify not only the consumer demographics but also trends based on date and time of day. This information can also be tied with other contextual information for improved product placement. The sensors can also calculate the average waiting time for customers as they checkout which, in turn, helps to improve customer service and operations.

In short, in-store computer vision sensing can help the retailer gain a better understanding of customer behavior leading to efficiencies in store operations, merchandizing, and marketing, thereby decreasing costs and improving customer loyalty and sales.

Moving Computer Vision from Research to Practice

While surveillance cameras and digital video streaming have become widespread, automatic understanding of the image/video content or "Computer Vision" is far from trivial! Thanks to billions of dollars and decades of research in universities, government and corporate laboratories, the field of Computer Vision has finally matured to the point where practical products are feasible. Moore's law, which states that the computing power will double every 18 months, has been true to its word and we are now entering an era where the real-time analysis of images and video streams is possible using standard off-the-shelf computing technologies as opposed to using high-end super computing technologies just a decade ago. Thus, what was only affordable to expensive defense and super computing applications is now available for the widespread consumer market.

Among the various research labs across the United States which focus on computer vision, the computer vision laboratory at Pennsylvania State University has obtained worldwide acclaim for achieving several remarkable breakthroughs in the automatic analysis of human behavior and composition from image sequences. These breakthroughs were a result of interdisciplinary research involving computer vision, artificial intelligence, pattern recognition, psychology, anthropology, cognitive science, and sociology, funded by large grants from the Department of Defense and National Science Foundation.

To capitalize on the commercialization of computer vision, the university spun out the research from the vision lab to form a new company, Advanced Interfaces (AI), in the first quarter of 2000, with the help of private investors.

AI's intelligence technologies work with surveillance cameras and other visual sensors to derive automatically the demographics and behavior of consumers in real-time. AI provides enterprise solutions to retail establishments and media networks to measure their customer composition and behavior. AI has acquired critical acclaim for its revolutionary technological applications and market potential with a growing base of prestigious customers and partners.

Market Potential for CRM Analytics

Retailers and manufacturers are increasing their investments in customer analytical technologies. Overall spending on CRM and sell-side e-commerce will reach $38B by 2005, at 29 percent CAGR, according to the AMR Research Report Customer Management Applications Report, 2000-2005. The market for analytical CRM is the fastest growing segment and expected to grow at twice the rate of the overall CRM market.

Implementing the solution

The solution architecture is simple. The video streams from in-store cameras and vision sensors are processed onsite using AI's patented real-time computer vision software and hardware methodologies. This allows the data compilation of each unique consumer, their gender, ethnicity, age-range, and other demographic and behavior parameters. The compiled data streams are then transmitted over a secure virtual private network to a central data warehouse facility at AI, where they are aggregated, further processed, analyzed and mined to extract the relevant information. The information that is derived is then fed into reports available to the corporate customers over a secure website. Additionally, popular market simulation software and decision support software can interface with AI data warehouses for valuable functionalities.

Other benefits - Targeted Media or "Narrowcasting"

Media outlets for advertising have become increasingly fragmented, with point-of-purchase (POP) advertising emerging as a powerful new medium to influence customers. However, these media networks, being deployed rapidly across major retail chains, are not measured media like print, television, radio and the Internet, causing inefficiencies to the advertiser and the media network. Thus, there is a need for measuring and understanding customer behavior in retail outlets in relation to new in-store media.

With technological advances, AI can leverage a retailers existing infrastructure of surveillance and deliver powerful analytical data for enhanced business value. Large retail chains including supermarkets, convenience stores, discount stores, department stores, and specialty retailers will benefit from these solutions.

Dr. Rajeev Sharma is the founder and CEO of Advanced Interfaces, a company that is pioneering breakthrough intelligence products to provide "behavior analytics."