Because it supplies many of the networks at stores, branch offices and remote sites, Cisco says it's in a good position to analyze data on site. For example, it is building data analytics capabilities into its ISR (Integrated Services Router) line, a series of systems for remote and branch networking that is available with a wide range of specialized modules. Other, much smaller and weather-hardened routers designed for use on rail cars or pipelines can do some analysis, too. Cisco's also building analytics into other products, including video security cameras that can decide whether a given piece of footage is interesting enough to save.
For examples of how video analytics can improve operations, Cisco described a deployment at an unnamed warehouse-type retail store. There, Cisco deployed systems that crunched video data locally to help with inventory tracking and staffing, Flannagan said.
The store trained video cameras on shelves to watch product availability more effectively than point-of-sale systems could, according to Cisco. Rather than relying on employees to check each product in and out of the department on the POS terminal, the new mechanism constantly watched the shelves themselves.
"From the video frames, we were able to determine when a shelf was empty and when a shelf had product," Flannagan said. If the boxes don't always get placed on the shelf correctly, machine learning can help the system identify the products anyway, according to Cisco. The cameras don't have to come from Cisco, Flannagan said.
The video analytics prevented errors stemming from things like returns, where an employee might take a returned, unopened item and put it back on the shelf before recording that action on the POS. The store's labor costs went down because staffers weren't wasting time restocking shelves that didn't need it, and revenue rose because not as many shoppers found accidentally empty shelves, according to Cisco.
Also using video, Cisco also used data about where people were in the store to predict when checkout lines would start to get crowded. The system learned to predict when a lot of shoppers were about to head for the checkout. For example, it found that customers tended to go to the frozen-food aisle right before they checked out. Based on that, the store could call more employees to the front of the store when the frozen section got crowded, Flannagan said.
Sign up for Computerworld eNewsletters.