Stream Optimization Vendors Make Big Claims About Reducing Bitrates, But Not Educating Market

There are a lot of product verticals within the streaming media industry, and one of the lesser known ones includes a small handful of vendors that are typically referred to as offering stream optimization technology. While many of them have very different solutions, the goal of all of them is the same. To reduce the size of video bitrates, without reducing quality. Vendors in the market include A2Z Logix, Beamr, Cinova, EuclidIQ, eyeIO, Faroudja Enterprises, InterDigital, Sigala and QuickFire Networks (just acquired by Facebook). Some of these vendors would take issue with me listing them next to others they feel don’t compete with them, but amongst  content owners, they are all thought of as offering ways to optimize video, even if many of them do it very differently. I don’t put them all in the same bucket, but many content owners do.

These compression technologies are already being applied to images, where companies like Yahoo and Netflix use solutions to make images load faster and save money. Over the past month, I’ve been looking deeper at some of the vendors in this space as it pertains to video compression, and have had conversations with a dozen OTT providers, asking them what they think of solutions currently on the market. What I’ve found is that there is a lot of confusion, mostly as a result of vendors not educating content owners and the industry on the technology, ROI, cost savings, impact on workflow amongst a host of other subjects.

Visiting the websites of many of the vendors provides little details on use cases, case studies, target demographics, cost savings, technical white papers and customers. Instead, many companies simply highlight how many patents they have, make claims of compression savings, have nice generic workflow diagrams, use lots of marketing buzzwords and all say how good their quality is. With some vendors, you can’t even tell from their website if they offer a service, a technology, a platform or a device. There are no reports, that I can find, from third-party companies, to back up claims that many of them make, which makes it hard for everyone to judge their solutions. I have yet to see any independent testing results, from any vendor (except one) that compares their technology to others in the market, or even to traditional H.264/HEVC encoding.

As I haven’t used any of these solutions myself, I talked to some of the biggest OTT providers, who deliver some of the largest volume of video on the web today. I won’t mention them by name since many gave me off-the-record comments or gave me feedback about specific vendors, but all of the ones I spoke to have evaluated or looked at stream optimization solutions in the market, in detail. Many of them have evaluated vendor’s products in their lab, under NDA. What I heard from nearly all of the content owners is that many of them view “some” of these solutions in a bad light. There is a stigma about them because some vendors make claims that simply aren’t accurate when the product is tested. Many content owners are still skeptical of the how companies can claim to reduce bitrates by 50%, yet still keep the same quality.

Also hurting this segment of the market was an article published a few years back that accused one of the vendors of selling “snake oil”. While that’s not a fair description of these services in the market as a whole, as one content owner said to me, “it helped create a stigma around the tech.” Yet some vendors are starting to show some signs of success and while this post isn’t about comparing technology from one provider to another, one or two vendors do seem to be rising above the noise, but it’s a challenge.

When it came to feedback from companies who have looked at stream optimization solutions, some told me that, “in real world testing, we didn’t see any compression decreases that we couldn’t reproduce with our own transcoding system.” Another major complaint by many is that the solutions “don’t run in real-time so it is a non-starter or way more expensive and latent for live encode.” Some vendors say they do offer real-time solutions, but content owners I spoke with said, “it didn’t work at scale” or that “the real-time functionality is only in beta.”

The following is a list of the most common feedback I got from OTT providers, about stream optimization solutions in the market. These don’t apply to every vendor, but there is a lot of crossover, especially when it comes to the impact on the video workflow:

  • lack of real time processing
  • adds more latency for live
  • it doesn’t neatly fit into a production workflow
  • forces the client device to work harder
  • inconsistent reduction in file size
  • bandwidth savings that don’t really exist
  • affects the battery life of mobile phones
  • makes the encode less desirable for encapsulated streaming
  • impacts other pieces of my workflow in a negative way
  • solution is simply too slow
  • didn’t see any compression decreases that we couldn’t reproduce with our own transcoding system
  • cost vs. ROI not clear, pricing too high

It should also be noted that one of, but not the only value proposition these vendors make, is the ability for content owners to save money on bandwidth. While that’s a good idea, the problem is that many content owners already expect to save on bandwidth, over time, simply by moving to HEVC. Content owners expect to see anywhere between 20%-40% compression savings, once HEVC takes hold. So that alone will save them on bandwidth, however HEVC adoption, at critical mass, is still years away. Vendors selling stream optimization products have to have more than just a bandwidth savings ROI, but that’s most of what they all pitch today. That is a hard sell and one that’s only going to get harder.

Another problem these vendors face is that while they are routinely talked about when the topic of 4K streaming comes us, and the need to deliver better quality with lower bitrates, there is no definition of what classifies 4K streaming. Bitrates can always be compressed, but with what tradeoff? What is considered acceptable? There is no standard and no spec, for 4K streaming, which is a big barrier to adoption, something that doesn’t help these vendors. One will tell you they can do 4K at 10Mbps, with no quality loss, another will say they can do it at 6Mbps. But those are big difference. Are they both considered 4K?. I don’t know the answer, no one does, since the industry hasn’t agreed on what 4K means for the web.

Even for a vendor who’s stuff works, they really have their work cut out for them. There is simply a lack of education in the market and vendors treat their technology as top secret, which doesn’t help the market grow and stifles education. Also missing is how one technology compares to others, what is/isn’t pre-processing and why you would want/not want one kind over the other etc.

Not to mention, no one knows what this stuff costs. Does it add 20% to a customer’s encoding costs but then reduces their delivery costs by 2x that? What’s the total cost to their workflow? What size company can use these solutions? When does it makes sense? How much traffic/content do you need to have? There are no studies are out there, that I can find, with real numbers, real ROI metrics, savings calculators etc. Most of these vendors don’t explain any of this on their websites, or say who their ideal customer is, which makes it hard for a content owner to know how big they need to be before it makes economic sense. This tech may very well work, be useful, have a good ROI etc. but to date, vendors have not shown/documented/proven that in the market, with any real education.

To be fair, some vendors have announced a few major customers, but we don’t know how these content owners are using it and how it’s been implemented. And the customer announcements are sparse. eyeIO announced Netflix as a customer in February of 2012, but I don’t know if they are still a customer or even using the technology anymore. Also, in the three years since eyeIO’s press release about Netflix, they have only announced one other customer in 2013, and none in 2014. Cinova, InterDigital, Faroudja Enterprises, Sigala and A2Z Logix, haven’t announced a single content owner customer via press release, that I can find. I know some of them do have a customer here or there, but nothing announced to the market. EuclidIQ has the best website in terms of the info they provide, but again, no customer details. Beamr has publicly announced M-GO and Crackle (Sony Pictures Entertainment) as customers and I hear has some others that have yet to be announced, but like the other vendors, their website provides almost no details.

As one content owner said to me about stream optimization technology, “In short its got just as many drawbacks as advantages best case.” That’s not good for a niche segment of the industry that is trying to grow and get adoption. The lack of education alone could kill this in the market before it even has a chance to try to grow.

  • Ulrich Grönewald

    Thanks Dan, excellent summary! Being in the video and audio codec business for 18 years, I can tell that “quality” is a very flexible term and that by tuning the test conditions in multiple ways (lowering certain restrictions, choosing specific source material, etc.) you can create yourself any advantage that you need for marketing purposes.

    Claims of “same quality” bandwidth reductions are easy to make and many may actually be true under certain conditions, but I could not have said it better than that: “In short its got just as many drawbacks as advantages best case.”

  • i’m noticing that the word compression is being used recently – where as really aren’t we talking conversion and encoding?

  • Ben Waggoner

    Yeah, quality has sort of a Heisenberg Uncertainty Principal factor. The more specific the definition/metric, the farther you get away from what the viewer actually experiences.

    UHD is also sort of an uncertain definition. It gets used interchangeably with 4K, which is properly 4096 wide, while UHD panels are 3840×2160. But there’s all that non-16:9 content, so you could define it as “at least 3840 pixels wide OR 2160 pixels tall.” In other aspects, it can be defined as >1920×1080. So what about 2560×1400? That requires UHD-capable decoder hardware and has UHD DRM restrictions typically. Would it be reasonable to call playback of a 2560×1088 widescreen movie on a 2560×1600 tablet screen “UHD?” And if not, what SHOULD it be called.

    Getting more specific as to stream optimizers, I think it’s helpful to just judge them as part of the encoding black box. With a given source, it takes X time to get Y result. All offline encoders have speed/quality knobs, and stream optimizers are effectively another knob, requiring an additional pass.

    Honestly, stream optimization technology in professional encoding should eventually become features of the encoder. If a company has a much improved way to get more bang for the bit, it is better integrated into the initial encode instead of encoding a whole file once and then trying to refine it. Refining it WHILE encoding would be faster, more efficient, and probably higher quality.

    I think the real sweet spot for stream optimizers is for user-generated content. Particularly mobile devices which have relatively weak real-time ASIC-ish CBR encoders. There will be a lot more bits to squeeze out of that than from a highly-tuned offline encoder.

  • Nora Georgieva

    Very insightful. Stream optimization being a relatively “fuzzy” vertical, due to all the reasons Dan points out, actually includes other types of solutions, not dealing with encoding, but rather with the way a stream is transported. Stream optimization aims to solve certain problems, and encoding is one way to go about it. Another way is to change the way a stream is delivered, or what peer-to-peer delivery solutions do. Interestingly, P2P solutions don’t share almost any of the characteristics outlined here. I’m curious, would you put them in this vertical, or you see them as a very different cup of tea? Disclosure: I do work at one such company, Viblast, and there are a number of others.

  • Purvin

    I agree with Ben. This technology should be part of encoders.