Do Video Analytics Work?

Manufacturers say video analytics work.  Customers and integrators often say they do not.  Are manufacturers simply lying? Are customers simply uninformed? Who is right and why does this gap exist?

Key Points

• What works is always relative to what one wants to do
• For some applications, some video analytics work well
• In general, most manufacturers overestimate how well analytics work relative to what customers require
  
• This results in low customer referrals and repeat business
• Market projections and expectations fail to be met

At the heart of this matter lies a deep truth that underlines and impacts the introduction of new technologies.

Manufacturers generally have a significantly lower standard for determining what works than customers or integrators.  This is not an accident yet it is generally not an issue of malice.  Most manufactures, especially at the senior management level, possess little domain knowledge, resulting in routine underestimation of the needs of their customers.

So when the question of "Does it work?" arise, manufacturers are generally willing to answer affirmitively if a product can withstand basic demonstrations and tests.  To the contrary, a security manager, with very detailed and specific circumstances to address, generally demands far higher and more particular standards.

Take for example, perimeter violation.  A manufacturer might assess the analytic to work if a person crosses the perimeter and an alarm is generated.  However, a security manager may demand that the analytic generate no more than 3 false alarms per day even if 10 cameras are being used.  The security manager may also demand if face a variety of weather and lighting conditions.

This contrast struck me as I interviewed more than 100 industry professionals over the last 3 months, including a dozen video analytic manufacturers.  The consistent theme I heard from the manufacturers were that their analytics did work (many went so far as to say their analytics required no configuration - a claim that rarely is achievable in the field).  The question then arose to me, "Am I being lied to or do these people truly believe this?"

I believe that most truly believed what they were saying but that their positions reflect a lack of experience in real world security operations.  The average manufacturer comes from a background in promotion (i.e., marketing) or development (i.e., CS or EE).  I rarely talked to individuals with any experience in security integration or security management.  Most companies do have people with security background but they are usually in field sales or services roles and have minor influence in product design decisions.

ASIS is soon and we will inevitably be faced with the question of "What video analytics do work?"  Bosch has already announced that it will again run its misleading video analytics demo that I critiqued a few months ago (true challenge for video analytics). Unfortunately, it seems that the show will encourage shallow and risky portrayals of what makes video analytics work.

Nonetheless, I am encouraged by an increasing number of public case studies of video analytics (including Phoenix IVS for public schools and ioimage's for the Israel legislature). If your analytics work, prove it with public case studies that security managers can verify.  This is a far healthier and better way for the industry to evolve than to continue to rely on shallow demos. Feel free to reference other case studies available online in the comments.