Aimetis CEO Marc Holtenhoff says users need to set the right expectations, form the right implementation teams, and choose proven applications in video analytics.
Achieving success with your intelligent video applications isn’t just about selecting the best products. Without advance planning, focusing on implementing the most appropriate applications, and evaluating deployment and management issues, you’re not likely to meet your goals. Here, then, are best practices for successfully deploying intelligent video applications.
Plan for success
First, establish an internal cross-functional team impacted by the new video system, comprised of IT, security and any other pertinent constituencies, such as operations.
Typically, IT implements, maintains and troubleshoots hardware/software/networking issues. Thus, it looks at the acquisition from a systems perspective, independent of the video applications. Meanwhile, security and other constituencies must communicate their feature and function needs with IT to ensure application needs and overall goals are met.
Once the internal team forms, it is paired with a team consisting of the vendor and integrator. The key to this partnership is trust. The customer needs to share information about its business requirements, note trouble points and discuss measurement criteria, timelines and budgets with the video manufacturer and systems integrator. The best way to ensure open and frequent communication is to schedule regular calls and meetings for brainstorming, problem solving, planning and status updates.
From this joint team, internal and vendor/system integrator groups should establish project management leads, who will give both sides a single voice to facilitate discussions and move the project forward.
Next, the joint team should develop a clear requirements statement. This is something that often does not receive the attention it deserves. First prioritize. Let’s say smash-and-grab crimes are up 20 percent, resulting in a $10 million increase in losses. If so, define the desired outcome based upon physical limitations and other barriers to success. Be wary of goals that seek perfect results. Too often organizations have been led to believe that a particular analytics application provides 99 percent accuracy when that number doesn’t reflect real-world conditions.
Expecting near perfect accuracy reflects a flawed understanding of analytics. For example, consider other helpful tools, such as email spam filters (perhaps 85 percent accurate) or even 5-day weather forecasts (maybe 75 percent). Despite their lack of perfection, we rely on both – and benefit from them. Likewise, analytics can provide tremendous benefits even if the results aren’t perfect.
As the user and as integrator, you will need to determine what level of performance is acceptable. For example, one retailer obtained 95 percent accuracy in its overhead people counting applications and noted with 80 percent accuracy how many customers lingered in front of specific retail end caps. In another case, a perimeter breach application that replaced motion detection reduced the number of false alarms from 300 to 4. Even though none of these applications was perfect, in each case the customer obtained genuine value.
The joint team should then look at system design, scope and support. Will your architecture be centralized, decentralized or even hosted? How many cameras will you place in how many sites? Will your desired solution scale across the enterprise? These are all important issues to review. Finally, figure out if the system fits within your existing video surveillance architecture and whether your IT and security departments can properly support it.
Select applications that deliver success
When evaluating applications, carefully select first those which you believe can be most successful. Consider both real-time alerts for prevention and forensics for recovery.
One proven real-time alert application is perimeter breach, which detects instances of people crossing a virtual, user-defined line. Such applications have proven to be more accurate than basic motion detection applications, which are prone to more false positives from small animals, blowing leaves, etc. Another area that has delivered strong results is people counting, which have found success in retail environments, schools, banks, recreation facilities, prisons, airports and other environments that require intelligent customer and operational insight.
Forensics can also provide tremendous value. Consider all the money spent on a security infrastructure, including surveillance cameras, recorders, video management and storage. Without analytics, many organizations have to manually search through hours or days of video to find a brief event. Even using motion search may only reduce footage by 10 percent or less. One our video system users figured it would have to spend 200 man hours per month per store to respond to events without analytics. Using analytics, which generally represents between 5 and 10 percent of the total surveillance investment, organizations can often look through a month of footage in an hour or less (or even a day in seconds) to uncover an event. Another way to look at this issue is to ask why companies bother to pay for storage systems if it isn’t practical for them to search through them.
The trend is clearly to shift from reactive to proactive alerts, updating control room video walls with intelligent qualified event-driven alerts. Further, organizations are now pushing intelligent-driven events to appropriate mobile responders. This reduces response times significantly and enables staff to examine more closely suspicious behavior that might lead to theft, property damage or safety issues.
Evaluate, deploy and manage
Next, create a proof of concept charter, which defines the objective, scope, all the set-up, testing/optimization and configuration work as well as the measurement method. This last item will ensure mutual understanding between in-house and vendor teams of what success looks like. For example, one objective may be to reduce false alarms by 50 percent, a goal which can be easily measured. Once an objective is achieved, the project leader will sign off on it and the team can discuss next steps.
At this point, you should be ready for deployment. But before beginning, make sure you take things slowly to achieve a series of small successes. Doing so will keep your constituencies and management supportive and the project on track. Think of it as crawl, then walk and then run.
Consider crawling first with one or just a few locations and with friendly users. Determine what works, what doesn’t and then tweak your plan accordingly. Assuming things go well, take a bigger step with an expanded user and/or geographical base. Based on your first experience, you should be able to automate installation and set-up (using bulk configuration and installation capabilities). During the final run phase, you’re ready for a full scale enterprise-wide roll out.
Once your video system is deployed, make sure that you train users, since the success of intelligent video is partly dependent upon the management and usability of the solution. Include both administrators and end users. Initially you’ll want hands-on training, but you can follow up with online distance learning and tutorial refresher courses. This is particularly important for users who may only use the analytics once per month or so.
When you are up and running, discuss with the vendor team how you can manage and upgrade the system as needed through software upgrades and onsite vendor support to augment your own IT department. Having experts in-house will ensure your system is properly supported and you can respond quickly to issues.
Implementing intelligent video applications isn’t trivial. So consider these best practices before starting your next project. By planning in advance, focusing on the most appropriate applications, and evaluating deployment and management issues, you’ll ensure a successful implementation and meet the needs of IT, security and other constituencies.
About the author: Marc Holtenhoff is CEO of video management and video analytics firm Aimetis, and is responsible for the organization's corporate strategy, financial performance and overall growth. Prior to Aimetis, Holtenhoff was the CEO of California-based 1GlobalPlace, an IT security company that he led from inception to its acquisition by VeriSign. Previously, he held a variety of enterprise sales and marketing management positions at the Xerox Corporation. Holtenhoff holds a master's degree in business administration and economics from the Copenhagen Business School in Denmark and a bachelor’s degree from Babson College in Wellesley, Mass.