I didn’t do predictions in 2011 and wasn’t going to in 2012, but two big trends seem pretty evident and worthy of prediction.
First, in 2012 you’re going to hear some major publishers claim they are making more revenue from online video than from online traditional (“banner”) advertising. It will probably start with pubs in the viral/comedy video space, or any type of quality content that can get viral or SEO distribution.
Second, in 2012 some other publishers, probably in the tech or blog space, will claim to get more traffic from mobile/tablet devices than traditional pc-like devices.
The first prediction is great news for pubs with video content. The second is dire news, unless monetization of mobile quickly catches up with established web-based deals.
This article originally appeared on the AppNexus Blog on August 1, 2011.
In my recent Ad Age piece on the disruptive nature of RTB technologies I took some shots at the rich media business. Specifically, I called it a “tech hairball”. Of all the points in the article, the rich media comments got the most feedback so I want a chance to explain exactly my criticism.
First, some brief background. The term “rich media” refers to display advertising creatives that utilize an ever-expanding collection of features like video, larger file size, social interactions, expansion beyond the banner slot, etc. Within the industry the term generally is associated with a set of vendors including PointRoll, MediaMind and DoubleClick (where I built and managed the rich media products). The fact that the whole business cannot be easily defined except in relation to the vendor offerings should raise eyebrows. Imagine if you couldn’t fully describe what a database was without reference to Oracle or Microsoft — that situation doesn’t exist in mature tech sectors, yet it is the case for the technology behind the most valuable digital ads running in display today.
At heart, the key issue that holds the rich media sector back is that the technical foundation of these ads remains immature and fragmented causing inefficiencies and complexities throughout the value chain. Rich media was invented by EyeBlaster and Unicast over ten years ago, yet virtually no standards have emerged to govern the delivery, reporting, or effectiveness of these creatives. In contrast, the in-stream video world has gone from inception to widespread adoption of the VAST and VPAID standards in less than five years. Mobile rich media is rapidly developing the ORMMA standard, potentially leaving us in a situation where it’s easier to develop cutting-edge mobile ads than browser display ads. Let’s examine the current state of rich media technology along the path of planning->creation->delivery->reporting->effectiveness and see how fragmentation continually limits the growth of the entire display business.
Planning
“Hey, let’s run a FatBoy on that placement!” For those not intimately aware, FatBoy is a trademark of PointRoll, and roughly corresponds to an expanding ad. I applaud PointRoll’s marketing strategy, which by all accounts has been hugely successful. But once again, imagine how silly this would sound if we were talking about databases in this manner, where the functionality was defined by a single vendor. For the sake of media planners and publisher campaign managers there needs to be a cross-vendor understanding of what the ads do. This lack of clarity on what rich media actually is, has practical effects because the same creative execution can be implemented very differently, and sometimes with different costs, across different vendors. The cost to the industry is wasted time, confusion, rework, etc.
Creative
Almost all rich media creative is based on Adobe’s Flash platform. Each vendor has built its own platform for extending Flash’s capabilities to achieve the various effects when the Flash creative runs in the browser. For example, in order to build an expanding creative, a vendor will build an interface for the Flash designer to expand the creative, and related functions to control the animation of the creative when expansion happens. But each vendor’s API is different, so while a DoubleClick creative might expand by the command conduit.expand(); the PointRoll creative will expand by pr.openPanel(1);. These are the obvious differences, but there are tons of subtle nuances you have to watch out for as a developer, like whether you need to manually stop your videos when an ad closes or not. Anyone in a creative department can attest to the fact that learning and optimizing against each vendor’s API is a massive pain in the neck, and causes the cost of creative development to go up and the quality to go down.
Delivery
There is a fundamental tension between the desire of the creative agency to push the envelope and the desire of the publisher to maintain a clean and orderly website with well-understood performance measures. Unfortunately, this tension is largely administered through ad hoc processes and undocumented controls. Publishers want to control which types of creatives can appear in which ad slots on their page. For a simple example, imagine a publisher wants to only allow 100K of file size per creative on the homepage, but up to 200K would be permissible on deeper content pages. There is simply no effective way to execute this strategy at scale. The current strategy is to load every creative in a browser, check the fiddler logs, and try to decipher how each vendor does “polite download” to limit customer impact. Most publishers still can’t either proactively control what shows on their page, or at the very least figure out precisely what’s being sent behind advertiser tags.
Let’s talk about iFrames. Publishers put their ad tags in iFrames to speed page download (the content of the iFrame loads asynchronously, thus prioritizing the rest of the page over the ad), and to protect their page from malicious creatives that might read the page content or otherwise affect the user experience. Rich media ads, though, need to affect the page content by expanding or otherwise giving the rich out-of-banner experience. The 10-year old solution to this problem is for each vendor to place its own customized “iFrame breakout” file on every publisher’s domain, thus bypassing the security of the iFrame. There’s so much wrong with this, I’m tempted to write a separate blog post. First, you’ve moved from an environment of security to one governed by trust. Second, the ad server needs to know, somehow, which domains have which breakout files. Third, there’s a huge barrier to entry for serving rich media creatives from new vendors, since they’ll have to do all the legwork to install the file on 5,000 sites. All of these problems get worse — much worse — when we move from a direct, sales dominant model to an exchange model since the algorithms will not reliably be able to determine which ad can run successfully on which site. The current plan from most vendors and exchange providers (AppNexus included) is to create lists of site-vendor combinations and target accordingly… Ugh.
In an ad network or exchange environment there are many additional data points we need to know in order to successfully deliver a rich media creative. For example, which direction should the expandable expand? The current solution is to ask the creative designer to develop the ad such that it works in multiple directions, but this is, in practice, inconsistently done. Another option is to choose the most common combinations (skyscrapers generally expand left, leaderboards generally expand down), and only support those in non-direct environments. But even if we knew which direction the ad slot could expand, we would have no way to programmatically know which direction the creative was built to expand, so we’re back to manually checking every creative in a browser. Obviously, that’s not ideal.
Reporting
Reporting is the most egregiously broken area for rich media, especially considering how important this is for evaluating the creative. Consider the clients’ perspective when buying expensive media with highly customized creative features like video or a game. The first question to the agency is “how did the ad perform?” Each vendor has created its own set of metrics to measure “interactivity,” “expansions” and other metrics. Unlike impressions and clicks, there are no standard IAB definitions for these metrics, how to calculate them, how to audit them, etc. As a result, the “interactivity rate” for a PointRoll creative will be different from the same rate on a DoubleClick creative, even if all other variables are the same. Consequently, the metric is devalued and the medium as a whole suffers. Next time you hear someone at a conference complain about “how confusing online metrics are,” think about this example.
Effectiveness
Building on the lack of standard metrics for rich media, very little research has been done to prove that these expensive, highly customized ads are actually more effective than more simple Flash creatives. Personally, I’m convinced they are – I just haven’t seen the data. At DoubleClick, we published a study showing correlations between rich media and Dynamic Logic brand studies, which was terrific. However for such a large portion of the display spend, it’s astounding to me how little justification has been done.
Recommendations
Here’s a partial list of what needs to happen to help rich media scale:
As someone with a history in this space, I take some personal responsibility for not pursuing these initiatives myself! Moving forward, I would be happy to help in the development of solutions. I look forward to hearing your feedback.
This article originally appeared on AdAge.com on July 22, 2011.
Real Time Bidding is the hottest technology in online advertising, and with good reason. By connecting buyers and sellers in real-time, on an impression basis, RTB has enabled an explosion of technology, experimentation, and measurement, resulting in improved results for buyers and yield for sellers. But even with the hype cycle in full force and growth in a hockey-stick patterns, the importance of RTB as a disruptive innovation is largely underestimated. Clayton Christensen first identified the concept of the disruptive innovation in the Innovator’s Dilemma. The basic idea is this: a new technology slowly undermines an existing, dominant technology, by starting out cheaper and “worse,” then slowly improving until it is a full replacement for the dominant one, but with newer, more flexible capabilities, and usually a lower cost basis. Classic examples of disruptive technologies include the PC (which disrupted mainframes and minicomputers) and desktop publishing (which disrupted the print industry). What started as “worse” will soon rule So while exchange technology was valuable to those who used it, it was (and still is) considered a low-end innovation, and much “worse” than traditional display buying. Certainly the interactive agencies handling brand campaigns, major product launches, and the like, weren’t getting excited about these buying opportunities a few years ago. In the second phase of ad exchanges, RTB-powered buying and selling grew like weeds, rapidly displacing most non-RTB, and simultaneously exchange buying displaced a significant portion of direct remnant sales to ad networks. With the leadership of the agency trading desks like Publicis’ Vivaki, ad agencies got the religion and started diverting real budgets from other indirect channels to RTB. This was very exciting! But even today the consensus remains that RTB is a great remnant technology, and will never seriously challenge the world of three-martini lunches, faxed IOs, relationships, and homepage takeovers. Simply put, it remains “worse” than the incumbent ways of doing business. Disruptive technology According to various studies the cost of buying online display advertising is roughly 2x to 10x that of other media. This overhead problem is blamed on the micro-targeted nature of online, where there are a near-infinite venues on which to buy, multiplied by numerous strategies of how to buy. This causes fragmentation of buying and selling, which, since the display market workflow was largely based on operational tactics borrowed from traditional media, doesn’t scale. The proposed solution to this problem has been to layer on interoperability layers like the IAB’s eBusiness initiative which automate the communication of the complexity across buyers and sellers. But what if a couple of years from now all the complexity just…goes…away? What if we don’t need to communicate the complexity. But rather we manage the complexity on an impression-by-impression basis using our own algorithms. If you just read that last sentence and said to yourself “no way, it will never work like that, it’s too complex” you just committed the first sin of evaluating a disruptive innovation. Here’s another example: rich media. The rich media business in display advertising is a poster child for dysfunction and lack of scale. Every vendor’s poorly-documented and relatively arbitrary product architecture is held together by bandages and bubble gum, yet these ads represent some of the most lucrative impressions online. This technology hairball simply won’t work in the RTB environment. So there are two options; either rich media ads won’t run in RTB (the current situation), or the more rigid and defined structure of the RTB environment will force standardization and declaration of rich media experiences, cutting through the gordian knot. Lower costs But what about homepage takeovers and martinis? Whither the media buyer? The fact that RTB is quickly taking over remnant inventory is, in fact, the first step in a major disruption to the entire online display ecosystem. The new technology stack will replace, in full, the current way of doing things, and sooner than many expect. The next time you hear a comment about RTB being only for the “least valuable” inventory, think twice about what the speaker is really saying, and try to imagine it playing out a little differently.
Ad exchanges were largely invented to enable ad networks to more efficiently and cheaply trade available impressions in order to meet performance goals. The word “performance” is important here — we’re talking about cost-per-click and cost-per-action deals and lead generation. This is the marketplace that, frankly, many of the mainstream members of the display ecosystem don’t consider very sexy or, even, in some cases, reputable. The liquidity this enabled was largely middlemen selling to middlemen, and the advantage of the technology was the avoidance of truly terrible operational options like “daisy-chaining” ad server tags and “pass-backs,” where impressions were passed back forward to an ad network, then passed back if unsold.
As ad exchanges went real-time, a whole new set of companies emerged to enable buying and selling in these environments. Products like Google’s RTB, and buying systems like those created by so-called demand-side platforms didn’t take the existing ad serving paradigm to heart, but rather created their interfaces from scratch for the new environment. Some interesting things happened in this build-out. A lot of the assumptions that are at the heart of the display ad technology went out the window. Instead of exchanging ad server tags by email and with manual checks for syntax, this process became an API. Instead of trusting that every buyer met the conflict rules for competitive advertising, the servers started making requests and responses with blacklists and advertiser declarations. These decisions, and many like them, are the building blocks on which this “worse” technology will challenge the old way of doing things to it’s core.
Most disruptive innovations exhibit lower long-term marginal costs than incumbents. Does RTB cost less than traditional ad serving? On the one hand, you have traditional ad servers, with fully-loaded cost structures designed for relatively simple decision making, which then have to rework their technology stack to the complexity of RTB. On the other hand, you have a brand new stack built from the ground up for RTB that then addresses the traditional market by operating on a vastly simpler decision making scale for selected clients. Which of these two stacks do you think will have even lower marginal costs as it scales to even larger volumes?
Homepages are always the argument people make with me about the limits of RTB. Let me be clear, I’m not suggesting that the human relationships, creativity, packaging, and overall high-touch associated with display advertising will be replaced by RTB. But, there’s no reason why RTB technologies couldn’t be used to execute a homepage campaign that was sold in the more traditional way. Gone will be the IOs and T&Cs, replaced with pre-defined private deal parameters. Gone will be the emailed tags, replaced by APIs. Gone will be the broken pages because of last minute trafficking errors, replaced with verified and audited creatives within a documented environment. And gone will be discrepancies and piddling fights over counting and data ownership, replaced with well-defined interfaces for counting and anonymizing data.
The Media Buyer’s role in RTB becomes much more important than ever before. But instead of evaluating sites and line-items on the budget, the media buyer needs to adapt to a world where each clients’ strategy may be wholly different, and where little counts other than the data. Brand buying moves from monthly panel-based statistics about top sites to historical views of past results and projections of those available through different buying strategies. Publishers need to step up with the data as well to differentiate and win the deal in an environment where supply is near unlimited.
Q&A: Ari Paparo of AppNexus on real-time bidding and yield optimization | Econsultancy
auto-decompile SWFs
Disruptive Innovation in Advertising: My Byline in Ad Age
For the past three months my Chrome browser has occasionally crapped out and stopped loading any pages at all while displaying the infamous “waiting for cache…” status message. The only fix that consistently works is to delete the cache from the tools menu, but this process can take 15 minutes to a half hour.
Some message boards recommend killing the cache files directly, but only for a PC. On a Mac, I couldn’t find anyone with useful information on how to permanently fix. Everyone seems to recommend deleting the cache located at ~Library/Application Support/Google/Chrome/Default/, but this directory doesn’t actually include the cache!
On a Mac, the solution is to delete all the files in the following directory:
~Library/Caches/Google/Chrome/Default/Cache
I just did it and it works like a charm. I’m writing this article in the hope that future sufferers can find the answer faster than I had to.
“Even with today’s “superphones” capable of video, local, and more, there are still significant barriers to executing a mobile strategy.” — My article at ClickZ on opportunities for ad networks in mobile: http://www.clickz.com/clickz/column/2076923/mobile-food-networks
The other day I tweeted about the most common come-on in online advertising, that integration of some new behavioral/widget/toolbar/bullshit service would be as simple as “one line of Javascript”. And it’s true, that you can implement a service that will deplete your business model, annoy your users, and destroy your user experience with jut one line of Javascript. With that in mind, I thought it was worth enumerating the ten biggest lies in Online Advertising. And not even with a slideshow — think of all the lost pageviews!
1. “Just One Line of Javascript”
Yes, with one line of javascript you can integrate virtually anything into your site. That’s not a benefit, it’s a threat. Just because it starts with <script src= doesn’t make it OK. Hey, just integrate this: <script src=”blow.my.head.off.js”>. Seriously dude, the script can a) capture all the data of your search referrers; b) cookie users for targeting based on your site’s profile; c) correlate those users to your referral URL, thereby profiling your site so it can be bought at a discount; d) read everything on your site, including confidential user data; e) add latency to your user experience; f) insert other javascript from other sources into your page; etc; etc. Think before you cut-and-paste.
2. Clicks and Clickthroughs Matter
This is only a lie to the extent that intelligent people benefit by continuing to believe it. Studies have shown that most people don’t click; that those who do are less desirable consumers; that most clicks are accidental; and that in any case clicks are very rare. Yet, clients and media planners continue to evaluate online advertising based on click-through-rates. There’s still a place for clicks on pure direct response offers or specific calls to action, but they aren’t a viable currency for generally evaluating online ads and should stop being treated as such.
3. Reach=cookies
Reach is defined as the number of people reached by an advertising campaign or media property. Cookies can measure the number of browsers exposed to an advertising campaign. Browser<>Person. Unless you have a mapping (panel, statistical, or otherwise) such that f(browser)=people then you can’t calculate reach. And just grossing up based on deleted cookies doesn’t qualify.
4. Television Dollars are Coming
Online advertising is a great medium that delivers quantifiable results to both brand and direct response advertisers. People are also using the Internet more every year, and watching less television. Therefore, Online Advertising will eventually take TV Dollars. No. This argument has more holes than Nielsen ratings. Online has done an amazing job growing at the expense of print, direct response, classified, and other media, and yet TV is bigger than ever. Until Online has the sight, sound, and motion of TV and ads are watched for more than a split second before the user browses away, it isn’t a viable alternative spending channel for TV dollars.
5. Display Needs to be More Like Search
Yes, search is incredible efficient, and measurable, and offers huge value to advertisers. And display is a pain in the ass, and fragmented, and difficult. But fixing display doesn’t mean making it more like search, unless by that you mean giving Google total control of the ecosystem. Oh whoops, that already happened. No, seriously, the great benefit and drawback to search is that every ad can be represented in 95 characters. In display, the whole point is to use creativity, color, context, and emotion to cut through the clutter. So while the mechanisms of buying and transacting display may become more like search, the creative itself will resist efforts at automation and that’s a positive.
6. Privacy is an Important Issue
Privacy in online advertising is an issue, to the extent that everyone is scared shitless that the government will do something stupid. But it’s a total non-issue in reality. There’s not a single documented case of a real-world harm coming from privacy-related abuses in online advertising. Fraud is an issue. Spam is an issue. Malware is an issue. Search results and social media contain a ton of privacy land mines. But the ability to know that an anonymous cookied browser represents is an “in-market car buyer” who recently visited KBB.com has yet to enter the realm of crimes against humanity.
7. “Faxing IOs”
This is my favorite since I think I personally used this lie hundreds of times. No one actually “faxes IOs” anymore, yet any vendor in the space who is talking up their crazy scheme to automate the workflow of the ecosystem likes to trot out this fallacy as a way of making us feel bad. We don’t fax IOs, we email PDFs of IOs.
8. Creative Burn-Out/Frequency Matters
This isn’t so much a lie as a rule of thumb that is outdated. Every agency out there makes sure to manage the frequency of exposure of their creative to avoid burn out. (Note, frequency is based on cookies, which are unreliable). The exact frequency is debatable, but is usually somewhere between 5 and 10 per cookie. Let’s do some math. In the TV world, an advertiser might try to limit frequency to 5 per household (of course, they can’t actually cap since it is broadcast). Five TV commercials = 150 seconds of ad exposure. The average web page is viewed for, I don’t know, maybe 5 seconds? So if you wanted the same exposure for web ads you would need a frequency of 30! And that assumes that a non-interactive, non-audio playing banner ad has the same effect per second of exposure as a TV ad blaring in an interruptive context. Personally, I think the ideal frequency for banner ads is near infinity.
9. Viewthrough Conversions with a 30-Day Lookback
A viewthrough conversion is when a user (really a cookie) buys something on your website and you give credit to the last impression seen by that user, within the “lookback window.” A lookback window is an amount of time in which you will give credit to that impression. The default lookback window most agencies use is 30 days. So if you saw a banner ad a month ago for a product you bought today, that banner can be given credit for influencing your purchase. Doesn’t make a whole lot of sense intuitively, and, to my knowledge, has no empirical research to support. The correct viewthrough lookback window is probably less than 2 hours.
10. Mac Users Don’t Matter
No one ever says that Mac users don’t matter. But they don’t. You know why? Because most Mac users use the default browser, Safari, to browse the web, and Safari, by default, does not allow any third-party cookies. So if you’re a) using cookies for reach/frequency calculations; b) using cookies to monitor and attribute conversions; c) using cookies to target users on your inventory; or d) using cookies to bid on impressions on exchanges, then you’re out of luck. Oh, also, Mac users are more educated and affluent.
Picture from construction site right outside ghetto liquor store.
Earlier today I tweeted my displeasure over the misleading headlines making the rounds that Facebook accounted for 31% of online display market share. A lot of fellow tweeters agreed or disagreed so I thought I would explain.
First, let me say this is not a disparagement of Facebook or their ad platform. While I was at Nielsen I worked closely with Facebook and there are some interesting studies showing the effectiveness of ads on the platform.
My objections is twofold:
What is Display, or How do we Define the Market
This is the crux of the issue. Is the market for “display” equivalent to the market for “ads that have images”? And if Facebook isn’t display, what is it?
From the perspective of all the current players in the display market, Facebook is a not part of the ecosystem. You can’t buy or sell ads on Facebook in the same ways as the rest of the market. You generally can’t use the same technology. You can’t use the same creative executions. It’s like we run an ice cream store, and when a new pizza shop opens across the street the press claims “pizza shop now half of calories in market.”
From the perspective of the buyers, there may be some overlap in budgets between Facebook and display, but there’s also tradeoffs in budget between display and search, email, video and every other “market” you might want to evaluate. So if you want to consider market share for “digital media buying” or “all advertising” then fine, go ahead and give Facebook the credit it deserves, but just assuming that all Facebook spending is competing against display is arbitrary and incorrect.
But it has Images!
Yes, Facebook ads generally include images. Does that make it display? Let’s imagine that Google announced tomorrow that for an extra $0.05 CPC they would allow advertisers to include a small image with every AdWords impression. Would all of AdWords suddenly become display? Or what if Facebook changed its mind and decided to not allow images anymore in their ad platform, would they suddenly be in a different market?
Then How Should we Measure Facebook?
There are plenty of very effective ways to measure Facebook’s progress against other media. Here’s a quick list:
But for Pete’s sake, don’t measure total number of impressions or compare to display impressions and pretend it’s meaningful.
I now support full marriage equality - Louis J. Marinelli -
I really came to understand that gays and lesbians were just real people who wanted to live real lives and be treated equally as opposed to, for example, wanting to destroy American culture. No, they didn’t want to destroy American culture, they wanted to openly particulate in it. I was well on my way to becoming a supporter of civil marriage equality.
“The Digital Video Committee is now taking another look at VAST and VPAID and, like Rocky Balboa about to get back in the ring, we are chasing a chicken around a yard. No.” — VAST and VPAID Update in Which We Chase a Chicken - IABlog
ormma -
ormma.org is an attempt to open source a standard for mobile rich media ads. If only the web were as standardized as mobile!
Bansky stencil of a worker cleaning up a previous generation’s street art.