LP Magazine Interview with Dan Cremins
Dan Cremins is director of product management with March Networks, a leading provider of video surveillance and video-driven business intelligence solutions used by retailers worldwide. He has more than sixteen years of experience in the video security industry. Prior to joining March Networks in 2009, he worked at Tyco/American Dynamics, where he held management positions in application engineering, product management, and business development. Cremins has spoken at leading security events and is a frequent contributor to security and retail publications on topics related to video-based applications.
Video analytics applications have evolved from earlier versions that produced less than spectacular results. When combined with other data sources in easy-to-use reporting tools, todayÕs analytics can help measurably reduce losses from theft and fraud. The challenge is defining up front what you want to achieve with your analytics, understanding what’s possible, and then setting up your system for success. To help find some of those answers, LP Magazine recently sat down with Dan Cremins, director of product management for March Networks.
What are some of the capabilities today’s analytics offer LP teams that previous generations did not?
Overall, the video analytics available today provide better accuracy compared to past generations. That’s not surprising, given advances in the technology and earlier experiences with analytics that didn’t perform to their promised levels, leaving many users feeling frustrated and manufacturers focused on improving performance. In that same vein of “lessons learned,” manufacturers also are careful to clearly explain what customers can expect in terms of accuracy. In some cases 85 percent accuracy might be perfectly acceptable for the application, while in other instances a retailer might opt to pay more for higher accuracy.
Faster and easier configuration is another characteristic of most of today’s video analytic products. An analytic that once might take an installer an hour or more to configure can now be set up in just minutes, thanks to innovative software features some manufacturers now build into their analytic products. For LP teams, that means time and costs saved on installation and less or no time spent beyond the initial setup to get camera calibrations working as they should.
Probably the most interesting contrast is how retailers are using analytics today. When the technology first appeared in the market more than a decade ago, it was primarily used for mission-critical security applications. For example, a retailer might use a tripwire analytic to trigger an alert if someone entered a high-value stockroom or a store after business hours.
While those security analytics are still being used and evolved, many of the new analytic offerings we’re seeing now have been developed to capture data for business trend analysis rather than for security. Pulling in other data sources, including point-of-sale (POS) transaction data and the synchronized surveillance video in easy-to-use reporting software is key to the value of these analytics for loss prevention.
A presence-detection analytic on its own is not going to provide much relevant information for LP, for example, but combined with transaction data and video, it enables LP investigators to start catching refunds where no customer is present. They can then use the same software to search for recurring incidents with the same employee or at the same store. Similarly, a people-counting analytic could help identify an instance of back-door theft by revealing an unusual number of entrances and exits in a given period.
What questions should we be asking systems integrators and solution providers regarding video analytics for LP?
The first thing I would recommend is ensuring that you have a very clear understanding of what your LP team wants to accomplish with the analytics being considered. Then have a conversation with your systems integrator, manufacturer, or both, so you know exactly what the analytic can realistically deliver.
|
A facial-recognition analytic might be ideal for certain applications, but if someone steals from one of your stores, itÕs still unlikely that you can take that image and check it against faces captured with surveillance in a crowd of shoppers in your other locations. If you want to receive an alert when someone is lingering for an unusual length of time in front of a high-value counter, which may be suspicious behavior or possibly a customer waiting for assistance, then a dwell-time analytic could be very effective.
You also want to understand where your analytics cameras will need to be placed to achieve the best results. Setting up an analytics camera for side-view capture, for example, when an overhead view is recommended can impact data accuracy significantly. I’d also advise LP professionals to consider what other views in their stores or restaurants need to be recorded for investigation and security purposes.
An analytics camera alone is not going to deliver complete coverage, and you will probably need to pair it with dedicated, high-resolution IP cameras over your POS terminals, perhaps a 360-degree camera for panoramic coverage, and others depending on the footprint and where surveillance cameras are already installed.
Remember that all of this information is important for LP professionals and systems integrators to understand. Your provider should be asking you how you’ll be using the analytics and discussing camera placement and other security and LP requirements, so they can guide you toward the best possible solution.
How do I make the best business case for analytics?
One of the most exciting things about business analytics is the ability to provide data intelligence that extends well beyond more typical security and LP applications. The same analytics that can alert LP teams to possible theft can also provide valuable insights for colleagues responsible for marketing, operations, and customer service, especially when combined with other data sources in user-friendly reporting software.
A people-counting analytic integrated with POS transaction data can provide important conversion-rate information that retailers can analyze and compare across stores and over time to identify areas for improvement. A dwell-time analytic can offer marketing groups detailed statistics on endcap display performance, while queue-length analytics can deliver relevant information on how long customers are waiting in line. If the answer is “too long,” then it’s easy to pull up the associated video to find out why, and then address the issue.
The ability to extend the use of surveillance video to other areas of the business by adopting analytics helps LP professionals build a compelling business case for the investment. Some may opt for a cost-sharing model and look to a system with flexible, user-based permissions. Other LP teams may continue to be the primary system users, while introducing the organization to a rich, new source of business intelligence.
However the model evolves, LP professionals are in an enviable position of being able to demonstrate how video surveillance combined with analytics, POS transactions, and other systems can provide retailers with a valuable, more complete view of their businesses and help increase profits.