Home » Data Protection » Network Security

Case Study: The ROI of Digital Video Surveillance

Allen Rude, security manager at Intel, invested more than four years in an ROI study to justify the cost of digital video surveillance

Shortlist in hand, Rude got some DVR systems from those vendors and called Intel Architecture Labs. He had the lab benchmark the DVRs versus the time-lapse VCRs he had deployed. This benchmarking, while rigorous, was purely technical. Rude wasn't interested in the real-world issues that would affect performance. He only wanted a lab-controlled horsepower comparison to see if the new technology, in and of itself, was better than the old. And it was.

Phase 2: 2001 to 2003—Pilot Systems and Productivity Benchmarking

Next, Rude added the real world to the mix. Ideally when he starts a pilot, Rude wants to visit other sites that have adopted the technology already and learn from them so that the pilot program will avoid simple mistakes that could derail the project. "It was difficult finding enterprise DVR customers at that time," Rude recalls. "But the gaming industry was adopting it. So we visited a lot of casinos."

Rude also had to choose between a greenfield and a legacy setting for the pilot. Testing the new system in new buildings would be easier. Piloting in older facilities would make it harder to deploy. Still, he chose older facilities rather than a greenfield, for two reasons. One, it made a more even comparison with the time-lapse VCR systems he had in other old facilities, and two, if his pilot proved the value of DVR technology there, he'd have done the hardest work already.

Rude had 808 cameras connected to dozens of DVRs at four sites, which were chosen for reasons both practical and arbitrary. Mainly, he wanted their sizes to roughly match the sizes of the sites with VCR systems, to which he would compare them.

Rude's team then began collecting the data for his productivity benchmark. He started with these calculations for each site. He based his metrics on the main expense (after capital expense, which will come up in Phase 3) he must pay for when he uses video surveillance: man-hours spent retrieving and reviewing footage.

Once he had this figure for each site, he averaged all the sites of each type (DVR or VCR) together to get one average monthly viewing time per camera.

The per camera per month numbers, however, are small, hard to apply and not very useful, because most deployments for Intel will number in the hundreds of cameras. So Rude made his data more real-world and executive-friendly by calculating the viewing time for 100 cameras over the course of a full year. He took his per camera per month number and multiplied it by 100 (for the camera total) and then by 12 (for a full year).