Thursday, October 02, 2008

HD ready?










My friends at Level(3) were kind enough to invite me along yesterday to their event, "HD Online: Can it be a Commercial Reality?" at BAFTA, which, perhaps unsurprisingly, has a very kick-ass presentation suite/screening room. The presentations were almost all uniformly interesting, with many common questions (how can ISPs be incentivized?, are ISP costs properly aligned with pricing?, can this stuff actually be monetized?) and a variety of answers/opinions. I can't go through each presentation in great detail, but will attempt to present the most interesting points. I found it to be a thoroughly informative and interesting day, and a great place to meet interesting, knowledgeable people. It always surprises me when I go to events of this nature and find no other financial people in the room.

Joe Trainor from Level(3) was the moderator for the morning session, and began with a couple of interesting points to highlight the fact that we are only at the beginning of a wild ride. New technologies initially start out aping their predecessors (early radio consisted of people reading the newspaper, early television consisted effectively of images of a radio announcer reading the news), before developing in unpredictable directions. And the gap between the arrival of a technology and its commercialization should not tempt us to be dismissive: John Logie Baird demonstrated television in 1926, but it was a full 15 years before the first TV ad was sold, to Bulova in 1941, for the princely sum of $9. The parallel with online video is that, while the first video stream rolled in 1993, it is only now, again 15 years later, that the industry is seriously trying to deal with the economics of this new distribution medium.

There followed two very interesting presentations, by Anthony Rose and Andy Quested from the BBC. Anthony Rose seemed to express fatigue with the iPlayer bandwidth debate, and stated his hope that the industry could now move on to consider issues around ISP incentivization and monetization. The iPlayer team has resisted the temptation to play out HD content so far, due to concerns that the experience would be unsatisfactory for many consumers, due to contention rates on DSL connections. (Unless I misunderstood his remarks during the Q&A panel later, I believe he said we should see some HD content on the iPlayer "this side of Christmas," probably encoded at around 4Mbps.) The iPlayer server farm, which apparently consists of 60 dual/quad core machines, transcodes content into "six or seven" codec/bitrate flavors, with the average in the 500 - 800kbps range (the latter being for H.264). The BBC is starting a trial of 1.5Mbps H.264 on the iPlayer to Virgin Media's 10,000 50Mbps trial customers in Ashford, Kent. The BBC's view is that the minimum threshold bitrate for HD is >3Mbps (though it's interesting to note that their "true" broadcast HD content goes out at 16Mbps or higher), which would be challenging for a lot of broadband connections and would risk high buffering levels. So there is a necessity to make sure the experience is not out of line with what HD TV viewers have come to expect. An essential ingredient in ensuring this would be an end-to-end dynamic adaptive bitrate system, but this is not a trivial exercise from a technology standpoint.

He made a case for the BBC's role in assembling the puzzle pieces for "others" (presumably ISPs) to build a business model around the iPlayer. He expressed an interest in working with ISPs to develop tiered service offerings to more closely align costs with revenues, as well as to cooperate on technology-based strategies to alleviate pressure on networks. Among these is a trial of Velocix network caching in three London suburbs (presumably this is with Virgin Media, but this was unclear).

Andy Quested, who is chief technologist on the HD side, gave an entertaining presentation, including a lot of pieces of sample footage encoded at different bitrates, in some cases shown in split-screen format, in order to demonstrate the variations in quality and the effects they have on different types of content (the flaws in lower resolution versions of a costume drama are much more tolerable than in a sports event, for example). He started out by observing that "HD" has developed into a brand phenomenon with many connotations and embedded expectations (he even had an ad for a pair of sunglasses which claimed to be HD - incidentally, this blog is coming to you in HD as well), which makes it a minefield if the user experience is not up to scratch. He also observed that the single greatest impediment to reaching true HD potential is the SCART cable, and that a large proportion of the complaints he receives on the broadcast HD service ultimately come down to cabling issues and incorrect surround sound configurations - yet typcially the unquestioning consumer only sees the experience as being less than what is promised.

Next came Kevin Baughan from Virgin Media, who highlighted that the Virgin TV iPlayer contributes 1/3 of all iPlayer sessions, at 11m sessions per month. There was other discussion about the "analogue dividend" and the eventual convergence of video and broadband bandwidth under DOCSIS 3.0, but the topic which really intrigued me was the question of storage vs. transmission. Virgin is experimenting with both edge/network caching, and in the Ashford 50Mbps trial area, has provisioned a 10Gbps link directly to Level(3), in an attempt to answer the question of whether storage trumps transmission, or the reverse. He conceded a fair number of unknowns around edge caching, such as predictability of demand and economics. I will be curious to learn what they determine in the process, but one intriguing idea he floated was that perhaps ISPs should be building their own internal CDN capabilities, striking deals with other CDNs and content players to mirror the most popular content.

There followed Phil Townend from ids, with some interesting observations on advertising. He cited forecasts of growth in online video streams in the UK from 250m per month in mid-2008 (of which the iPlayer only accounted for 1.2%) to over 650m per moonth by 2010, with ad revenues reaching £200m by 2012. Forecasts aside, there is current evidence of some serious traction in the market: Virgin Media has apparently sold 40m pre-roll ad streams so far in 2008, and PHD (part of Omnicom) now includes a tranche of online video ads in every TV campaign they run, apparently as an experiment. He also pointed to large opportunities to open up new industry verticals - for example FMCG giants like Unilever and P&G only spend around 2% of their total advertising budgets in online, but he believes online video may be just the sort of environment to pull them in. Another interesting observation was that the industry needs to move beyond pre-roll video ads, as they are too disruptive and test viewer patience, to the extent that a 30-second pre-roll generates abandonment rates of 40%, vs. only 15% for a 10 - 15-second clip, which has now unsurprisingly become the dominant format. He went on to show some examples of overlay ads and "in-skin" ads/branded media players. In the case of the overlay example (Kraft Foods), there was a standard pre-roll, then an overlay over the core "program" (itself an infomercial), which, when clicked, pauses the core program and presents another clip which gives more specific product information. His point here was that one piece of content can now be sold as three separate avails, effectively trebling the CPM. This was a key area of interest in the panel Q&A which followed, given that ad revenues in an HD environment will need to offset the transport costs of the content, which are likely to be at least a factor of four times those of standard definition video.

The speakers were then reassembled onstage for a panel exchange and Q&A. There was a fair amount of repetition here, but a few very interesting tidbits came out. On the inevitable question of the incentive for ISPs if CPMs remain static but delivery costs quadruple, Andy Quested answered that maybe the real question is whether audiences will watch non-HD material in future, pointing to HD displacement of SD equivalent channels in viewing patterns in the US. Anthony Rose chimed in intriguingly, alluding to a TiVo-like "season-ticket" predictive functionality in the works for the iPlayer, which would download content overnight, when most networks are relatively quiet. The next time the user visits the iPlayer site and clicks on this content, it will begin playing, but from the hard drive rather than as a stream. Kevin Baughan proposed that there is actually a need to move beyond the CPM to some other more suitable measure, such as share of attention, or duration of session. If the average duration of a single view on YouTube is under three minutes, then should advertising around this be priced in the same way as a session viewing long-format HD content (the BBC guys pointed out that the average iPlayer view is 22 minutes, with the average session consisting of two programs, i.e., 45 minutes, with very high completion rates)? The conclusion was probably not, but the industry is obviously going to take some time to work out appropriate metrics and structures.

After a top-notch lunch (can food be in HD?), Griff Parry, head of on-demand services at Sky, took a somewhat more cautious tone. His main messages seemed to be that the industry should concentrate on doing SD on the web properly before attempting HD, and that linear TV is far from dead. In support of the latter point, he cited viewing figures which show that, even in Sky+ households, 79% of viewing is still live/linear, despite 1.4 billion instances of time-shifted viewing on Sky+ last year. His take on HD seemed to be that it is currently a niche product for Sky, though it is emerging as a customer acquisition tool. As for HD online as a commercial proposition, he didn't rule out the prospect that it might eventually form a new channel for Sky subscriptions independent of satellite, but the overall impression was that this was a dark horse.

Microsoft followed with what was, in my opinion, a fairly dull presentation - the only relevant point of which I can recall being, that in the Democratic National Convention streaming project, jointly provided by Microsoft, Level(3) and Move Networks, the average viewing session time was a whopping 107 minutes.

Next came John Edwards, founder and CEO of Move Networks, in what was, for my money, the most interesting presentation of the day. Here we have an over-the-top video technology with 50m definable endpoints, 12 petabytes of traffic per month, average viewing durations over 60 minutes per session, a unique dynamic adaptive streaming capability coupled with a one-time "simulcoding" approach, edge caching solutions, time-based and geo-based ad targeting - in short, what more could we ask for? One really interesting illustration of the power of Move's platform, which tied in nicely with earlier discussions of the need to move beyond pre-roll ads, was the "Toyota Intermission," which consisted of a talking torso with five video frames above her head and to the left side of the frame. Four of the five frames seemed to relate to specific aspects of the car model on display, while the fifth frame was an avail for a local Toyota dealership, generated by (apparently) a commercial relationship and the geo-targeting capability of the platform.

Sadly, due to a previous commitment, I had to leave early on during the final presentation, which was by Steve Allison, Technology Evangelist for Adobe, so the only bits I caught which were of interest were that Flash 9 is at >90% penetration of global PCs after less than one year, and that Flash has shifted 159 petabytes of data so far this year.

Overall, this was a great event, though I left it with far more questions than answers. I noted, during informal chats with other delegates, that there was a concern with the "real economy" which threw the reality of the proceedings into question. I can't offer much solace there, except to note that some of the big "bubble era" stock victims of the 1929 crash were in fact radio companies, then a highly speculative segment, which ten years later developed into a hugely profitable and influential industry, which is still with us as a powerful force today, though unrecognizable from its orignal form.

UPDATE: A Palladium Club Mega-Uber HD Value Reader points me to a highly relevant and interesting paper from the excellent Dave Clark, which should be of interest to anyone interested in this topic.

No comments: