Some video analytics math
I didn't catch up with Cernium CEO Craig Chambers at ISC West so I caught up with him yesterday on the phone. As part of a pretty wide ranging conversation (which matched up pretty well with the video analytics discussion we had at TechSec), he offered up one piece of math that called into question the tactic of selling video analytics licenses. Because that's basically the debate in the video analytics market: Do you sell products that have video analytics as part of their features, or do you sell video analytics licenses to other people who make products and make your margin on the sale of those licenses? Is video analytics a product unto itself, or is video analytics just a feature of another product? You've heard this before, I'm sure. The license guys point to Dolby - they seem to be doing okay just selling their technology to other guys who make products. The product guys say, "you don't see people making a living selling chemistry. But chemistry makes other products possible. Analytics are just a technology that makes other things possible." But here's Craig Chambers' math: Say there are 10 million cameras sold worldwide in 2010 (maybe it's a little more or less, it doesn't really matter, you'll see). Say 10 percent of them have video analytics associated with them (that's probably a high number, considering only maybe 10 percent of all the cameras sold will be IP cameras, but that's made up for by DVRs with analytics and encoders with analytics, etc., so it's 10 percent of all "channels"). That leaves you with 1 million cameras that can have a video analytics software license associated with them. So, how much does that video analytics license cost, wholesale? What's Pelco pay for an OV license on a Sarix camera, say? Chambers postulates that it's about $10. While many video analytics companies drew up business plans that assumed licenses would retain values near $100 for a while, he says the market, flooded by analytics vendors, has crashed down into the single digits for some licenses. But go ahead and say you can get $20 a license for your embedded analytics. What is there to the market? Maybe $20 million in total potential revenue, split up between how many companies? Go ahead and double every piece of the equation - you're still talking about less than $100 million worldwide no matter how you figure it. Chambers argues that unless you've got a product to sell, wrapped around the analytics (Cernium's Archerfish and CheckVideo, VideoIQ's ICVR, ioimage's little red boxes, even Aimetis' VMS software), there's just no way to create a scalable company with analytics. There's not enough total money, no matter how good the margin is on algorithms. Maybe this isn't a new concept. Plenty of people have made the argument to me that analytics are a feature, not a product, but that math sort of brought the argument home for me. Will the number of cameras explode to 100 million, with half of them having analytics on board, thereby creating a much bigger pot? Seems unlikely. Maybe a company like BRS Labs can show so much value with their algorithms that they can continue to get hundreds of dollars a channel for their analytics, but that's going to take a lot of differentiation to show the return on that investment. And, yes, I know I wrote a story referencing VB/Research's numbers on the growth of the market, where they took the total value of the products that have the analytics on board and made that represent the total market for analytics, but, yeah, I'm coming around to the idea that that's a little like saying the market for the guys who make the sensors that tell my car if I've got my seatbelt on or not is represented by the entire number for auto sales worldwide.