Does Lower Expectations Solve Video Analytics Problems?

An increasingly popular but dangerous trend is the argument that expectations for video analytics need to be lowered.  The thinking is that since video analytics have had problems in the field, if we lower the demands on what video analytics need to do, customers will be more satisfied. This report explains why this is wishful but deeply problematic thinking.

What are Expectations Today?

My experience talking with hundreds of people on analytics is that most of them expect video analytics to work consistently with almost no false alerts and simple setup.

Aren't some Expectations Unrealistic?

Sure. One time a VC asked me if we could do an analytic that detected blinking so that a hostage could secretly trigger an alert. Another time a Program Manager at a defense contractor asked me if we could send an alert if a person got on a train with a package but did not have the package when he exited the train at another station.

In my experience, these are exceptions.  I cannot remember a security manager asking for a science fiction video analytic setup. They consistenly ask for and expect performance to meet their operational needs.

What would we Lower Expectations To?

A number of leading experts suggest that our expectations should be that video analytics will have moderate false alerts, require significant on-site setup and have significant limitations on use.

The Impact of Lower Expectations

While this may be the truth for many video analytic offerings, such expectations would kill the video analytics industry.  These lower expectations would make video analytics useless to almost all security managers.  The capital and ongoing operational cost to handle such a limited performance solution would be very difficult to justify.

The Solution is Making It Work

Sometimes, lowering expectations just is not tenable. You cannot ask parents to simply lower their expectations when their child is failing out of 8th grade. Likewise, you cannot expect security managers to simply accept performance that would undermine their operations.

The only solution for the industry is to make it work and to prove that it works. Part of our challenge is that it's so hard to know if and how well any video analytic works. If you are a vendor and are interested in doing so, let's talk. I would be happy to do so and share the results with the community. Anyone who has first hand experience, please share your results.