Online video advertising takeaways November 15, 2007Posted by jeremyliew in ad networks, advertising, branding, business models, video.
I moderated a panel today at NewTeeVee Live, which was one of the best conferences that I’ve attended this year. Kudos to Om and Liz for putting together a great show.
The four panelists were all CEOs of video ad networks. Jayant Kadambi of YuMe, Tod Sacerdoti of Brightroll, Matt Sanchez of VideoEgg and Matt Wasserlauf of Broadband Enterprises did a great job of discussing some of the challenges and opportunities in the industry.
Here are some takeaways from the discussion:
Online video is a brand building medium, bought on a CPM basis.
Ad buys today are coming from three buckets; TV budgets (the big opportunity, but further away), online budgets and experimental budgets. Video content coming from TV and other traditional video producers is most likely to win buys from TV budgets. Native online video content still coming from experimental budgets.
To get to a scalable online video ad business, the key barriers are (i) standardization of ad units, (ii) standardization of ad serving infrastructure and (iii) standardization of measurement/metrics.
There is tension between highly effective preroll ads and user friendly overlay ads (with a user initiated play), but these are the two most common ad units and most likely to become a standard. In-banner video is also popular.
Big Media companies prefer point solutions for online video so that they can sell and control their own inventory; smaller online content owners want a packaged solution that delivers all elements of the value chain (ad sales, ad targeting, ad inventory management, ad serving, ad measurement) from a single provider. There is increasing pressure to create standards in ad serving from the large media companies so that multiple sales forces can sell ads without having to implement multiple serving solutions.
Advertisers are buying content adjacencies today as a proxy to demographic targeting, but they would prefer to buy demos directly where they can
The four panelists really engaged; it was a pleasure to listen in to their conversation. You can read the transcript of the panel here.
Online ads targeted by offline data October 18, 2007Posted by jeremyliew in ad networks, advertising.
Wednesday’s WSJ has a fascinating article about how Acxiom mines offline data to target online ads. It’s the most targeted and data rich approach to targeting online ads that I’ve heard of or seen and I’m very surprised that it didn’t get more blogosphere coverage.
Acxiom’s new service, Relevance-X, goes further, drawing on the company’s database of 133 million households to determine which ads to show. Acxiom’s consumer database includes information gleaned from sources such as public real-estate and motor-vehicle records, surveys and warrantee cards consumers fill out. Estimates of annual income, marital status, average ages of kids, home ownership and property value, educational level and travel histories are also available.
The company classifies each U.S. household into 70 clusters based, it says, “on that household’s specific consumer and demographic characteristics, including shopping, media, lifestyle and attitudinal information.” Clusters range from “Married Sophisticates” to “Penny Pinchers.”
Acxiom contracts with Web sites that collect consumer addresses, such as online retailers and those offering sweepstakes and surveys. In a blink, Acxiom looks up the people who provide their addresses in its database, matches them with their demographic and lifestyle clusters and places “cookies,” or small pieces of tracking data, on their computer hard drives.
When those people visit Acxiom partner Web sites in the future, Acxiom can read cluster codes embedded in the cookies and use them to pick which ads to show. The company doesn’t disclose the sites that carry such targeted ads, but says they reach 60% of U.S. Internet users.
That allows a company selling an expensive antiwrinkle cream, for example, to contract with Acxiom to display its ads to affluent women 40 years or older in the “Skyboxes and Suburbans” or “Summit Estates” clusters.
Acxiom says that there are no privacy concerns because only gender, zip and the segment that a person belongs to are stored in the cookie and used to target, and segments are all at least 1m people. Personally, I don’t see why there should be any privacy concerns for online targeting since this same data has been used to target offline advertising for a long time. But this is exactly what caused Doubleclick’s acquisition of Abacus Direct to come under huge scrutiny in 2000, and led to the subsequent spin off of Abacus.
This data could have a significant positive effect on industry wide CPMs if its targeting can really improve the effectiveness of online advertising.
Will contextual advertising work for online video? August 27, 2007Posted by jeremyliew in ad networks, advertising, contextual targeting, video.
I posted last week about why I though that Google’s new overlay advertising product would be good for the whole industry; Gootube has both the volume of inventory and the advertiser relationships to make the overlay a standard ad unit.
The other notable thing about Google’s new ad product is how the ads are being targeted, or rather how they are NOT being targeted. The New York Times quotes Eileen Norton, Google’s Director for Media Platforms:
Ms. Naughton also said advertisers would be able to take aim at specific channels and genres, as well as demographic profiles, geography and hour of the day.
What is notably missing from this list, especially from Google, is contextual targeting. I wonder if this suggests that contextual targeting is not as important for online video ads as it is for text link ads.
Online, Google’s adsense has been the premier form of contextual targeting, and it is primarily about direct response.
Television advertising is primarily about branding. Direct Response TV (infomercials) make up a very small fraction of TV advertising and they typically run in latenight and overnight time segments when both ratings and ad prices are low.
The question is whether online video advertising will look more like online, or more like TV.
For long form video online, it seems less likely that contextual advertising will be a good match. Long form video is more of a “lean back” experience, where the viewer is less likely to click on any ad, even a highly contextual one. That makes it hard for direct response advertising to work.
Short form video online may be more promising as viewers may be more willing to click away. People from online video analytics comapnies tell me that less than 50% of online video streams are watched to the end, with the bulk of the drop off occurring in the first 20% of the stream.
When you combine this with the fact that both Youtube and VideoEgg are seeing click through rates on their overlay ad unit 5-10x higher than typical online banner ad click through rates (according to the NY Times article), it seems more possible that direct response advertising will work for short form video online.
But two factors complicate this situation. The first is that neither Youtube nor VideoEgg are actually using contextual targeting today. The second is that the current advertiser base for VideoEgg appears to skew heavily towards “cool” entertainment ads (gaming, movies, TV and music), or at least so it appears from their sample advertisers. The same is true of the few Youtube ads that I have seen “in the wild”.
These early adopters may have more compelling video ads that are not as representative of the mass market – it may be easier to get someone to click away to watch a Superbad trailer than an ad for Tide, even a good one.
If indeed it is true that targeting against channels, genres and demographics is sufficient for video advertising, then this is great news for online video startups. Google accuracy at contextual targeting its text ads benefits greatly from the vast volume of ads that it serves, and from its very low cost compute infrastructure. Targeting against channels, genres and demographics requires a lot less volume and a lot less computation, which levels the playing field substantially.
I’d be interested to hear what readers think about whether contextual targeting will be the way forward for online video advertising.
To our disappointment, there has been extremely limited uptake by the advertising community around [overlays]. There are a lot of factors behind this limited uptake, including:
– the advertising community buying video have been very focused on leveraging existing creative and buying patterns in the online video space
– most content publishers and media owners have been focused on getting the ‘basics’ up and running, and also responding to the RFPs from marketers and advertisers, which are almost 100% focused on basic short-form video commercials
– for premium brands and content, the basic pre-roll and companion banners are yielding extremely attractive CPMs and there is little evidence that :15 ads have any negative impact on end-user viewership behavior — in fact, our own metrics show that sites that run without any ads, and then introduce :15 pre-rolls and banners achieve identical usage and performance (e.g. no drop-off in users because of ads) on their content.
Nonetheless, we remain very bullish about ‘composite’ video advertising formats that combine overlays and unique and non-intrusive calls to action with deeper interactive marketing experiences. We’ve been pushing this for years and only now are starting to see the publishers and media owners that we work with begin to take an interest in these formats. I believe this is because we’re now entering a phase where content companies are looking at ways to maximize yield and revenue within their content, and they are introducing more mid and long-form content which require, by economic necessity, a different suite of formats to deliver a good user experience.
Jeremy’s experience is not surprising. As I have said in the past, new forms of advertising are hard. They take longer to catch on then you expect. Until standards emerge, it can be difficult to cross over from “early adopter” advertisers who are willing to experiment, into the mass market of advertisers. If the media buyer at the agency doesn’t see your sort of advertising as a line item, she can’t allocate you part of the ad spend.
That being said, Youtube’s entry into the market is a game changer. With Youtube representing 50% more market share than ALL other online video sites combined (according to Hitwise), and with Google’s existing relationships with advertisers, they have both the volume and the connections to be able to create a standard. And that is great news for VideoEgg, Brightcove, AdBrite, and all the other online video ad networks. Online Google/Youtube can create the standard that the industry needs to be able to really grow into scale.
Hidden traffic drivers at top tier sites August 14, 2007Posted by jeremyliew in ad networks, advertising, business models, Consumer internet, Internet, web 2.0.
While Glam is more like a vertical ad network (which it readily admits to), there are many other well known sites that derive meaningful amounts of their Comscore traffic from unexpected sources. Among Comscore’s top 100 web properties, Ask, CNet, the New York Times, Move.com and iVillage all also generate the majority of their network pageviews from sites other than their namesake:
The non-namesake traffic was mostly driven through acquisition, although in some cases (e.g. Zwinky) the growth was organic. In many other examples though, acquisitions have been absorbed into the URL structure of the acquiring company. Yahoo for example acquired Launch (now music.yahoo.com), Hotjobs (now hotjobs.yahoo.com) Geocites (now geocities.yahoo.com), among others, all of which now roll up under the Yahoo.com URL.
Whether ad networks or acquisitions, Comscore’s “media property” reporting often includes a lot more than the namesake URL in the rollup. While this can come as a surprise to the unwary, it is no surprise to the people that matter – the people who are buying online advertising. As one media buyer commented on the Techcrunch article about Glam (abridged quote):
As an online media buyer, perhaps I have a different opinion then most of the outsiders commenting on the sidelines. I use comScore and NetRatings on a daily basis for planning media spends targeting large audiences online… The only way to measure the audience that any large media property reaches is through a panel based media measurement tool like comScore… It does not matter if they own a site or have a partnership with them.
This stuff is no big secret.
The Prisoner’s Dilemma in online advertising August 1, 2007Posted by jeremyliew in ad networks, advertising, Consumer internet, economics, video, web 2.0, widgets.
I posted previously about how increased innovation in online advertising is driving up costs. Online media companies would generally prefer more standarization and less customization in online advertising; this makes their processes more scalable and keeps their costs down. However, they face a prisoner’s dilemma situation that has made it hard to drive standardization as an industry.
The prisoner’s dilemma is a staple of game theory classes. Wikipedia summarizes the problem as follows:
Two suspects, A and B, are arrested by the police. The police have insufficient evidence for a conviction, and, having separated both prisoners, visit each of them to offer the same deal: if one testifies for the prosecution against the other and the other remains silent, the betrayer goes free and the silent accomplice receives the full 10-year sentence. If both stay silent, both prisoners are sentenced to only six months in jail for a minor charge. If each betrays the other, each receives a five-year sentence. Each prisoner must make the choice of whether to betray the other or to remain silent. However, neither prisoner knows for sure what choice the other prisoner will make. So this dilemma poses the question: How should the prisoners act?
Classic game theory predicts that in a single instance of the game, the dominant strategy is to betray your accomplice. However, if the game is repeated, the best strategy for rational players repeatedly interacting for indefinitely long games can lead to sustaining the cooperative outcome.
The Wikipedia article cites several real world examples of the prisoner’s dilemma, including one involving cigarette advertising.
When cigarette advertising [on TV and radio] was legal in the United States, competing cigarette manufacturers had to decide how much money to spend on advertising … cigarette manufacturers endorsed the creation of laws banning cigarette advertising [on TV and radio], understanding that this would reduce costs and increase profits across the industry.
While not advocating that we use cigarette companies as a role model, I believe that the online advertising industry currently faces a similar opportunity to reduce costs and increase profits over the issue of increasing customization in online advertising that I posted about last week.
So how does this relate to the prisoners dilemma? Rather than the police asking suspects to confess, advertisers are asking online media companies for costly custom advertising. If one media company is willing to customize and its competitor isn’t, then the customizing company is more likely to win the deal.
But if both companies customize then creative and production costs go up while the size of the ad spend does not. More money is spent on creating the campaign, and less goes to buying media. Thus both media companies suffer.
If neither company customizes, then less money is spent on creative and more goes to buying media and filling the online media companies’ coffers.
To make this situation more complicated, there aren’t just two prisoners who need to cooperate, but rather many online media companies. With many players, it can be very hard to drive towards a cooperative outcome.
For media companies, the “cooperation” case means adhering to a set of standards in creative format. While this doesn’t eliminate the costs of creative, it does at least set boundaries to help control creative costs.
While these standards exist in banner advertising, (728×90, 300x 250, 160×600 etc), they do not yet exist in other, newer forms of online advertising (including social media marketing, widget marketing, online video marketing, and casual immersive world marketing). But through the IAB, we saw standards eventually emerge in banner advertising, and hopefully we will see the IAB and other standards bodies (perhaps the newly formed Widget Marketing Association?) help set standards within the newer forms of online advertising as well.
This is a necessary but not sufficient condition for the industry to converge to a stable “cooperative” equilibrium in this version of the prisoner’s dilemma. I’ve campaigned for standards in social network advertising before.
AOL Buys Tacoda – some thoughts on why July 24, 2007Posted by jeremyliew in ad networks, advertising, portals.
1 comment so far
The NY Post breaks the news that AOL has bought Tacoda, the leading behavioral targeting ad network. I wrote a guest piece on Venturebeat that went up last night about vertically targeted content sites and how they can add value to both synthetic channels and behavioral networks, topics I have blogged about here in the past.
AOL has made a lot of very interesting advertising acquisitions recently, starting with Advertising.com a few years ago, and more recently Third Screen Media, Adtech AG, Lightning Cast and now Tacoda (and excluding their attempt to buy Tradedoubler). This is why I commented on Yahoo’s acquisition of Right Media that they were not following Google’s acquisition of Doubleclick, but rather following AOL. All portals are losing market share of pageviews and need to increase their reach offsite to maintain their advertising effectiveness. Congrats to Mike Kelly, Jorge Espinel and the team
Today’s WSJ has an excellent article on behavioral targeting. It details Pepsi’s launch of Aquafina Alive,their new low cal vitamin enhanced water. The campaign was backed by an online campaign through Tacoda and targeted to people who had previously visited “healthy lifestyles” websites.
The result? Pepsi recorded a threefold increase in the number of people clicking on its Aquafina Alive ads compared with previous campaigns. “We’ve never been able to get to this level of granularity,” says John Vail, director of the interactive marketing group at Pepsi-Cola North America.
Brand advertisers like Pepsi (you’re not buying the water online after all!) having this sort of success is a strong indicator of the growing importance of ad networks.
Last week the FTC announced that they would investigate Microsoft‘s acquisition of aQuantive and Yahoo‘s acquisition of Right Media, adding them to their ongoing investigation of Google‘s acquisition of Doubleclick. Although the American Association of Advertising Agencies and the Association of National Advertisers asked for antitrust review, I think that they have little to fear and much to benefit.
First, let me give some context on ad sales. As I guy who has “carried a bag” myself, I believe that sales cycles are directly proportional to the complexity of the sales message, and RPMs inversely proportional.
A simple sales message results in a short sales cycle and high RPMs. Take automotive advertising; imagine you’re selling online advertising to Toyota.
Autoblog has the next easiest sales proposition. Autoblog has amassed an audience of auto enthusiasts. While they may not all be in-market car buyers, they have a natural affinity to cars, and selling ads against this vertically targeted audience is a relatively simple proposition with relatively high RPMs.
The Washington Post sells its demographic. It argues that its readership skews higher income and hence is more likely to buy a Lexus. Its a pitch that has worked in print and on TV for years, and works online as well, although not quite as well. Sales cycles will be reasonable, and RPMs lower than for Autoblog.
Many large sites or networks sell reach. The pitch goes something like “Lots of people visit us, and some of them might want to buy a car, right?” Its a tougher sale, and even when successful, it generally results in CPC campaigns that mitigate the advertiser’s risk, often resulting in low effective RPMs.
So what does this mean for the networks and the “fattening long tail”? As I’ve posted about in the past, I’m a fan of “synthetic channels“, or vertically targeted ad networks, that can take advantage of the first two of these categories of sales.
But many sites generate pageviews that are unable to take advantage of these sales pitches. They are untargeted, and as such, are likely to be soldat a deep discount as “run of site” or “remnant” inventory, if they are sold at all.
This is where ad networks who apply behavioral targeting can really have a big impact. eMarketer predicts that behavioral targeting will increase by 7x over the next four years.
A large network can anonymously track a user as they move around the internet, recognizing them to be the same person when they show up at different sites across the network. They then use this data to target ads more effectively on “lower value” sites, thereby increasing the value of the ad inventory. So for example, a user who had previously visited Autoblog gets a Lexus ad when they show up on a MySpace page, even though the Myspace page has nothing to do with cars. This has become standard procedure at many of the big networks.
As we saw from the Aquafina example, behavioral targeting works for advertisers.
The key question is how does the incremental value (the increased CPM between the Lexus ad and a “run of network” ad) get split between the three constituents in the relationship; Autoblog, MySpace and the network. This is partly the outcome of supply and demand. Because of the massive surplus of “broad reach” inventory online, the bulk of the value goes to the network and Autoblog.
The network gains disproportionate value from having Autoblog in its network because it can now make a lot of its “broad reach” inventory more valuable. However, in a world with multiple networks, a site that ads “behavioral information” to a network can take that data advantage with it to any other network.
So sites with the opportunity for endemic advertisers have an advantage, not just as standalone sites, but also as members of ad networks.
Its not the advertisers that need to worry about consolidation of ad networks, but sites with content that can be used for behavioral targeting.
Ad Networks: Synthetic channels June 11, 2007Posted by jeremyliew in ad networks, advertising, business models, Consumer internet, Internet, start-up, startups.
One of the hallmarks of the last few years on the internet has been the growing length of the “long tail”. Compete released some data last year showing that its panel was visiting 77% more websites than it did five years ago:
Interestingly enough, Compete also released data showing that the “head” of the internet was growing in size:
Together, this suggests that there are now many new sites getting only moderate traffic. While these sites may never grow big enough to become public companies, they are very likely getting to a scale where they can break even. I spoke on this topic at the web 2.0 expo where my presentation analyzed how big sites needed to get to hit both of these goals.
Many of the smaller ad supported sites turn to ad networks for monetization. This trend is being matched by advertisers embracing the channel. A recent report by Collective Media found that:
* 66% of advertisers plan to increase their usage of ad networks in 2007
* 88% of respondents planning to use online ad networks in 2007 (up from 77% in 2006).
* 57% of respondents believed how an ad network targets audiences was the #1 differentiating factor between networks
* Reach (at 52%) and Efficiency (at 66%) were still the key drivers for why agencies/advertisers include ad networks on the buy
I believe that among the biggest beneficiaries of these trends have been the content specific ad networks. With more advertising buying on networks, and with audience targeting being the #1 differentiator, networks that can offer extremely targeted audiences focused where an advertiser is endemic are hard to beat. As I’ve posted about in the past, having endemic advertisers makes for higher RPMs, and hence a smaller level of overall traffic scale to get to high revenues. At AOL, the channels with endemic advertisers always got the highest CPMs and sell through rates, and content specific ad networks are essentially creating synthetic channels.
Many ad networks do contextual targeting in their efforts to get endemic advertisers next to content (with Google and Yahoo being the most prominent). Others use behavioral targeting. Both of these approaches have been effective in lifting RPMs, but both require a leap of faith from the advertiser that the “black box” truly works. To mitigate this risk, contextual networks usually have CPC or CPA based pricing models. However, these models don’t capture all the value of a branding campaign, which can only be fully priced by a CPM model. This leaves some value on the table. Many endemic advertisers are not looking just for “in market” buyers who are looking to make a purchase decision imminently. They are also looking to build brand awareness to influence future purchase decisions.
Synthetic channels, like the channels on the big portals, have an advantage in this respect. By guaranteeing that all sites in their network are about a single topic, they can aggregate a critical mass in traffic while still enjoying endemic site RPMs. This is, in a sense, a “hack” to true contextual targeting, but it has the advantage of being simple to understand and hence simple to sell to advertisers.
One example is Jumpstart, a synthetic channel reaching 5m UU/month and focused on the auto industry. It was bought by Hachette Filipacche (publisher of Car & Driver and Road & Track) in April of this year for up to $110m.
Another example is Glam, which started life as an site focused on fashion, but quickly morphed itself into a synthetic channel focused on “Women: Fashion and Lifestyle” and reaching over 12m UU/month. It claims to be the “fastest growing web property in 2007“.
A third example is the Health Central Network, a synthetic channel focused on medical information and tools. Many of Health Central Network’s sites are actually owned and operated by the company as it rolled up small health content sites during the internet bust.
I think we’ll see more synthetic channels emerge, focused on the high ad spend categories:
Note that automotive, retail and medicine, the content targets of the three examples above, are the three of the top four advertiser categories. I’m sure this is not an accident! Are readers aware of other synthetic channels?