|
Published:
4/12/2013
Thanks Joel, great point!
|
|
Published:
3/4/2013
Michael Hubbard, thanks!
The last time I calculated impression volume I came up with 5 Trillion monthly impressions across the whole online display space. This is based on private conversations with folks over drinks who own the serving infrastructure at the largest publishers and ad servers, plus some public data. Note that the Comscore and Nielsen numbers are radically off when compared to ad server numbers. I found that when I compared numbers I had access to and numbers I was hearing from other publishers to the numbers from Comscore and Nielsen the differential was pretty proportional. So I'm confident we're well over 5 Trillion display ads a month. Cause last time I did this was 2 years ago. ;)
|
|
Published:
1/10/2013
Thanks Raj! I can't wait for this to happen in all media.
|
|
Published:
9/13/2012
@Ben, thanks for your comments.
The trend is to make products more and more differentiated, not more general. Even products that traditionally were for everyone are becoming more segmented, and as a Search guy you probably operate this way yourself.
Take toothpaste for example. We all (hopefully) use it. But the trend is to specialize and differentiate. Whitening toothpaste with peroxide and baking soda aimed at women who are engaged and getting married in six months.
Yes there are broad reach products, but usually the more effective marketing approach is to hone the message toward a target audience. In theory the use of DMPs allows even tighter refinement, but the creative isn't generally going to be individualized, and the most you can target on at scale is 3 (maybe 4) targeting parameters today.
Even in paid search, you refine the audience you are reaching by selecting keywords and queries, and possibly vary the message based on that - and possibly even audience attributes. If your goal was just to reach anyone, why use keywords at all? Its because that person fits your target audience of people searching for that product. And the longer and more refined the query, the better, right?
Using personas is looking at people more like individuals. More than the alternatives, I think.
|
|
Published:
5/21/2012
Happy to help get the word out, Peter! This is important work!
|
|
Published:
10/18/2011
John, Thanks so much for the comment! I'm a bit confused how increasing the efficiency of media buys and making display more brand friendly are diametrically opposed. They may be more orthogonal than diametrically opposed. :)
|
|
Published:
4/22/2011
Hi "Gertrude",
You wrote:
"...you say "The publisher shows them ads, and frequently requires that the consumer register or subscribe (regardless of if this is a free or paid subscription) and hand over some data to be used to better sell ads to advertisers." and on theother that you do so on a site which demands that one registers in order to post comments!! "
I think perhaps you didn't catch a key point of my article. I don't have a problem with publishers requiring registration or subscriptions. Nor do I have a problem with publishers targeting their visitors with ads based on profiles of their behavior. This first party tracking and targeting is defensible in my opinion. My concern is actually about 3rd parties - companies that are not publishers or advertisers who are wiring up hundreds or thousands of websites with tracking pixels in order to track activity across many sites, and then target ads to the users. I don't even really care about the idea of targeting ads based on third party data - I care about those 3rd parties tracking people without their permission. Because the people they're tracking don't really get any value out of it - and there are real privacy risks that are being perpetrated against those users without any permission.
As far as your example goes - you're complaining about a request to register with a publisher in order to make a comment. iMedia is a professional trade publication - and fosters a pretty direct dialog between professionals. While you can choose to create a fake persona to enable anonymous commenting, which you did, I think there's real value to a direct, persistent conversation.
And on the side of the Quid-Pro-Quo I argue exists between reader and publisher.... You've gotten some value out of the article I wrote - it seems. At least enough to want to comment on it. And only when attempting to comment did you realize the requirement to register.
In this case, iMedia didn't require you to register to read that content, just to comment on it. Which is even extra value. I think for this kind of publication, they'd be justified in requiring registration to even read the article. But I chose them as a publisher specifically because they foster a good community, their conferences are incredibly high quality, their editorial staff is dedicated and ethical, and because they don't block easy access to my content - fostering a big audience to read my stuff.
Eric
|
|
Published:
4/20/2011
James - thanks for your comments!
George - Thanks! I believe that 3rd party cookies (and data sources) will be relegated to a marginal role over time. And that Publishers and Advertisers (each with their own 1st party data relationships with users) will be able to bridge to fix a lot of these issues. So there will be a drop in scale initially, but over time we'll catch back up to where we are today. And perhaps the role of the 3rd Party data providers will become more consultative to the first parties than it is today.
Any movement toward cracking users privacy once they've opted out is obviously not a good idea and should be avoided. The real win would be if someone can convince a large percentage of users to share their data.
|
|
Published:
1/20/2011
John,
Sorry I missed this until now. I think we're in agreement overall. The key point is that if publishers don't expose their 'premium' inventory in a non-blind fashion, TechCrunch (in your example below) won't get the benefit of the advertiser's knowledge of the venue. If TechCrunch let's the data about content association be part of the bidding rules, then they will get higher bids. They also should set a floor price on this inventory so that they're not getting channel conflict with their sales force.
I think we do need to be careful about the term 'yield' in this space. "Certain advertisers and products experience no incremental yield improvement by associating their advertising with a particular content category." Advertisers don't typically think in terms of yield unless they're old-school DR (offline) background buyers. Yield typically refers in online display to the revenue that a publisher gets from selling the inventory - higher being better for them. Advertisers think in terms of ROI typically - and how they define it is broadly misunderstood. Many brands are looking at 'cost to reach' their desired audience as their ROI metric. Many online DR folks are looking at CPA.
Ultimately we as an industry need to be enabling large brands to spend money efficiently in order to reach their desired audience. This efficiency is critically important to our industry - it's not cost effective to buy targeted reach in our industry today for a big brand. Offline is much more efficient for them to buy. We need to get better at this.
Eric
|
|
Published:
9/14/2010
Shachar,
I'm not clear on what you just said, let me make sure I understood it.
Are you agreeing with me that people don't want to be followed around the web by re-targeting solutions, or are you agreeing that people find this kind of solution creepy instead of valuable? Or are you saying that you agree that we're likely to get legislated against like the telemarketing industry in the US?
I'm not sure how Opt Out will solve any of these problems for consumers.
I'm also not saying I have a good answer to what should be done. I think that Opt In legislation looks more and more likely - and frankly the more folks in our industry pound the Opt Out drum when these issues come up, the more likely legislation seems to me.
I'm not a fan of legislation - see my last article for a long discussion of this - I think it will stifle the growth of this industry at a time when the industry is just figuring itself out. But I don't think we're making a compelling case for why legislation shouldn't be passed yet.
|
|
Published:
8/12/2010
Thanks for the thoughts EricW (and great to see you here!) I don't think the state of the art in general is going to proceed to the point where consumers will see a large enough percentage of their advertising as targeted ads any time soon. The simple fact is that even if the technology is ready (which it isn't) there needs to be enough ad inventory that is purchased this way before enough ads will be able to be delivered to be noticable. And then, what's the outcome? Consumers won't really notice anything if this works properly - they'll just find the ads more useful. This will be a subtle change - one that is positive - but one that won't stand out for them (unless it is actually so targeted that it's creepy.)
I don't think we can argue for this that way - it's going to take too long to get stabilized for things to actually feel better to consumers before we end up having to answer legislators.
Let me ask you this... Do the ads that ad agencies create and media agencies place on web sites work? Do they increase sales? Is there actual proof of this? I think the answer is yes - because I've seen enough case studies to know that advertising works. And companies typically don't waste money on things that don't work.
So if advertising works, then my argument about economic benefit should be relatively self-evident. What needs to be proven is that targeted advertising works better, is more efficient, and more effective. I've seen enough case studies on this as well that I'm confident that the answer is yes. So what is it that we need to do in order to push this new meme out to the world?
Eric Picard
|
|
Published:
5/20/2010
Troy,
You make a point that the problem is 'not enough advertisers' participating to spread the demand across lots of impressions, and if the >400K advertisers were bidding this wouldn't be a problem. It isn't so simple. First - getting that number of bidders is really not so simple, and isn't necessarily a logical progression as it could seem on the surface.
The problem you're describing is bid density across broad coverage of impressions. As I explained in my article above - the problem we have is that the number of display advertisers is radically smaller than paid search advertisers. This leads to problems of asymmetric bidding and low bid density, and of coverage.
There are many reasons that it is unlikely for the >400K paid search advertisers to be replicated in display - so this isn't a simple problem to solve. Even with the proof point of contextual this isn't really happening. Some of that comes from goals - advertisers are willing to expend the effort to get at impressions that sit very far down the purchase funnel.
Even small advertisers in vertical niches will do so on Google - e.g. A rare bookseller will spend an hour a week tweaking their keywords and ads in Google because they might sell one or two books a month that way - which probably is a good ROI for them. That wouldn't be the case for display ads unless we had fantastic targeting data for rare book collectors, and even then - we wouldn't necessarily know that they were looking to buy like we do when they type in a title of a book and the keywords rare, vintage or collectible.
The vast (vast vast vast) majority of display impressions do not sit in that location on the purchase funnel. This means that the number of advertisers willing to advertise in those locations is simply lower. Also there are significant creative production issues that limit access to a class of advertisers who can easily type in a text ad. I make a point in the article that without automated creative customization against targeting criteria we won't see bidding that rightfully takes advantage of the rich targeting criteria. But this is a different problem from getting a display ad created at all. Some companies have entered the scene trying to solve this - and we'll see if they get traction.
If we could get every Paid Search advertiser to buy display ads, increasing coverage in the auction, and bid density on more impressions, it wouldn't solve the problem. When you're bidding on audience characteristics, rather than contextual keywords or content associations you still end up with a rather limited set of supply. e.g. Zales, Home Depot, Best Buy and Whole Foods all most desire the exact same people as customers. Once you are spreading your audience characteristics across the 5-6K advertisers that buy display, you just don't get the coverage you're describing from 400k advertisers bidding on 'end of funnel' keywords.
|
|
Published:
5/19/2010
On DSPs 'running out' of RTB inventory. Keep in mind - that they are bidding only on impressions that match their campaign goals. If there are no impressions that match their campaign goals, then they're running out of inventory. e.g. There is more demand than supply on those impressions. This leads to higher yield on these competitive impressions, raising publisher yield. But that doesn't mean that publishers will always 'clear' all inventory at any price. Just because a publisher has inventory doesn't mean there's a market for it. Even at zero cost.
The strange thing is that there is an expectation in online display that publishers should be able to liquidate most of their inventory. In no other media in history has that been the case. And we've seen this flood of cheap low value inventory cause significant harm to the overall average CPM and likely yield altogether.
Publishers would be much better off designing their pages to look good with or without ads, and simply not showing ads that are incredibly low value. I'm not sure why page views = ad inventory in every case. There aren't ads on every page of every magazine. We're overloading our audience.
|
|
Published:
5/13/2010
Alan -there are many benefits. Simple ones include:
All DSPs let advertisers integrate their proprietary customer data into the mix, letting the DSP include it in the bidding process. This lets the advertiser keep that data proprietary without exposing to a third party.
In general it puts the buyer (and media agencies) back into a high value position. The roles will evolve - requiring more technical buyers/planners - but this model is widely viewed by many in agency leadership as the way they save their businesses. It solidifies the position of the media profession and enables it to evolve forward into the 21st century.
Because every impression is evaluated in real time, the amount of granularity in the optimization for ROI is much greater - and lets the buyer tweak bids for each impression (programmatically - you don't want to rely on humans for this. :)
As Rob (and others) have correctly pointed out - this becomes a technically significant process, and at high scale it will require significant engineering prowess. Many of the DSPs have never built anything with the kind of scale and sophistication that will ultimately be required. But they have time to figure it out.
|
|
Published:
5/13/2010
Hi Rob - a few responses to your comments:
"in practice (effectiveness of RTB has) not yet proven in any meaningful way."
I disagree. I've talked to dozens of agencies and advertisers who say differently. They're more than convinced - they're investing heavily.
On your comments about floor prices - it's important not to use the 'blunt instrument' example of reserved media sales between humans. Hulu is a clear example of a blunt instrument at play - but I think they're probably yielding higher overall for not selling off remnant inventory at any available price. It protects their user experience (although as a heavy Hulu consumer I'd like to see a wider variety of PSAs) by keeping low end DR creative out of the mix.
Watching Hulu at any time of day never feels like watching 'tail' cable at 1AM. So better user experience leads to higher viewership - and the yield from remnant is so low to begin with that it's better to leave it unsold in many cases. I've long been a proponent of premium advertisers kicking their remnant crack habit. TV limits the amount of remnant to about 10% - so should online.
|
|
Published:
5/12/2010
Hi Zach - you're absolutely right that my example is too simple. As I said in the article - I'm making an overly simple example to illustrate what could happen. I opted for something readers would understand over realistically driving the message home. I created an artificially perfect impression and only put 4 bidders on the impression.
On the agency issue - I think you're missing something. In the old world, where humans negotiated every buy - you would rarely find buys that specified more than a few parameters. So you'd have one advertiser reserve a million impressions of males in an age group. One advertiser reserve a million impressions in the Chicago DMA. One advertiser reserve a million impressions of Auto Shoppers. One advertiser reserve a million impressions of New Parents.
The 'old school' publisher delivery engines do very basic allocation of impressions against campaigns. Since the campaigns are all reserved - they can't really optimize for yield and for hitting individual campaign delivery goals. In one of these systems - it's completely possible that this perfect impression that is worth so much to some advertisers would be allocated to a low-yielding campaign that is 'starving' more than the others - just in order to hit their impression goals. That's incredibly wasteful, but the real-world we live in today.
So what I'm saying is - the agencies today will have multiple campaigns for single-dimesion targets that compete for the same impression inventory on the same publishers. In a world where that exact methodology moves to auctions in 'proxy mode' rather than RTB - they'd compete with each other for the same impressions driving up yield - because they're likely all interested in similar audience characteristics - or users who overlap in characterstics (e.g. Male 34 years old who is an Auto Shopper and a New Parent.) Keep in mind that people with income over $200K in the US are only 1% of the population. They're all high value targets for almost every marketer.
In RTB where the same agency manages a portfolio of campaigns - they absolutely could offer only one bid for an impression rather than one for each campaign for each advertiser they're representing. This is pretty basic game theory - and will ultimately be a selling point for agencies. Regardless of whether in the past you put Pampers, Ford and Zales in different categories - they're definitely going to compete over audiences in this new world.
|
|
Published:
3/11/2010
Hi Bryan - great to see you're following here!
This article was really written for marketers more than IT Pros, so I didn't cover all aspects of this or give it deep analysis. Clearly beyond a certain scale, you can save costs by building out your own infrastructure. But you might find if you work with your finance partners, there are some accounting rules that could come into play that change the game a bit on things like amortization - but I'll leave that to the tax pros.
I think calling Cloud Services another brand of Managed Services missed the value of cloud services a bit. If you're just using Cloud as a managed services replacement, you probably could get a bit more value from it than that. But for this audience, I don't think we'd do them much of a service by debating this.
I think the elasticity of Azure or EC2 are among the most valuable pieces that change the game more than just a bit. That ability to let the cloud handle the load and not particularly worrying about having enough servers in the farm. And you should check out SQL Azure and especially something they're doing under 'codename Dallas' as it relates to data and being able to build apps on top of huge available data feeds.
Just rebuilding ad serving (granted - an example I used) in the cloud is a basic application of cloud computing. The data assets and letting enterprises that don't handle massive amounts of data in their current IT infrastucture begin pushing that scalability out to the cloud is really where marketing will see its greatest value.
On the cost side - Bluestreak's a bit mature (and has a large fixed cost sunk into datacenters already) for rebuilding everything in the cloud. But you might consider using the Cloud for specific elements of your business - e.g. Putting a bunch of your reporting in the cloud, or if you get around to building a real-time bidding engine (becoming a DSP) you may want to build that new business on Cloud infrastructure to defray the startup costs.
Fun times!
|
|
Published:
1/4/2010
Steven, thanks for your comments. A couple of quick responses:
You wrote: Rick Boyce didn't sell the AT&T ad, Jane Metcalfe did (Rick joined HotWired months later).
Thanks for this correction! This industry is full of word-of-mouth lore, and that kind of correction is super helpful for setting the record straight.
You wrote a bunch about how sales people didn't spec out ad platforms, and comment about GRPs/TRPs not working back in those days as penetration was not very broad.
My point isn't that Sales People speced out the ad platforms. My point is that the requirements that were gathered by those creating the first generation ad platforms were gathered from sales people who were setting 'go-to-market' objectives for online display advertising that were appropriate for an emerging media, but that were not appropriate for a mature media. So while the best way to launch a new media type is to sell broad long-term sponsorships, once a media achieves a mass market penetration and wants to operate at scale, it needs to evolve. Unfortunately all the software written back in those early days locked us into an operating model that is designed for a small scale media type. So breaking out of that operating model requires software changes that are not on any company's roadmaps for their ad platforms.
|
|
Published:
11/24/2009
Hi Doug - thanks for the comments. The one adjustment I'd make to your interpretation is that I believe we will see high value audiences monetize well regardless of where the ad is delivered. The question will be what happens to the impressions that are not identifiable as high value consumers. I suggest that overall, impressions that are not able to be targeted against high value audiences will drop in price. But that the price for premium publishers (where at least the context of the ad and the quality of the environment are not in question) will drop marginally - while the already low value impressions on non-premium publishers will be valueless and essentially free. We'll see some DR spending on those pages and those that perform well will continue to show ads. Keeping in mind that if the cost is low enough, performance can be quite poor and still be revenue positive.
|
|
Published:
11/11/2009
It occurs to me that it could come across as if I don't value sales people, or media buyers. This couldn't be further from the truth. But in this new model their role will change significantly. Sales will be far more evangelical, planning will be more highly technical.
|
|
Published:
10/8/2009
Sorry about that Jeff, it was clearer when I first wrote it, but after several edits, I inadvertntly edited out the position you actually held. Apologies.
|
|
Published:
9/11/2009
Carmen - I'm not seeing these Botox ads, how strange! Wait... Aaaahhh... :)
Cyhyoung - The point I am making is that large advertisers want to spend large amounts of money to reach large audiences. The amount of work needing to be done to reach an audience of 50 people just isn't cost effective. But for a small business, it can be absolutely cost effective and very high ROI. It all depends on the ROI an advertiser sees.
For Ford, it probably wouldn't make sense to have someone at their agency spend 4 hours a week trying to reach a very small (even a valuable) micro-segment of 50 people. They need to sell thousands of cars efficiently with ads at the National level. But it makes perfect sense for a local Ford dealership to spend 4 hours a week trying to get at those 50 people.
|
|
Published:
7/14/2009
Ari,
I'm not sure you understood my article based on your response. I did NOT say that any Rich Media company had dropped the ball. I'm saying that as an industry, we have dropped the ball. Every single basic Flash ad that has been built and delivered over every ad serving system out there could have all the functionality I suggested in my article.
LinkStorm is interesting, and I wish you well.
|
|
Published:
6/24/2009
Andrew - thanks for your feedback.
You said: TV buyers must constantly steward their accounts and negotiare makegood & bonus packages for underdelivery.
While certainly this is necessary in both online and offline media, and I may have slightly overstated the point - the amount of work done in online display (or for that matter, even paid search) is orders of magnitude higher than traditional media. This is especially true once the campaign has been trafficked, but is also true for the research, planning, and buying phases (to a somewhat lesser extent.)
While in this article I brushed past this point quickly - that's because I covered it much more extensively in the article below:
http://www.imediaconnection.com/content/23023.asp
|
|
Published:
5/14/2009
"online advertising" - I'm not sure that you're commenting at all on my article. Looks to me like spam.
|
|
Published:
5/14/2009
Optimizing campaigns is an incredibly sophisticated problem to solve. While it can be automated, the automation is in no way a 'one size fits all'. Determining a 'starting point' for optimization is by no means something that can be done completely by computers. The role of media planner / buyer is not going anywhere - the role will become more sophisticated, more technical over time.
An unsophisticated model only works for unsophisticated customers. Big advertisers are not unsophisticated, nor are their agencies. Big publishers are not unsophisticated. In no way does anything I've written here presuppose a 'dumbing down' of the market. In fact, if we change the metaphors and mechanisms, we can 'supercharge' the industry while supporting significantly more sophistication.
I argue that we've limited sophistication by overloading participants in way too much granularity of control as the baseline assumption.
|
|
Published:
4/8/2009
It occurs to me that I didn't mention the fact that during a downturn, many advertisers will 'rejigger' their spending and weight toward DR models. That's why we see the decline in display revenues so significantly during a downturn - same amount of inventory gets sold, but more of it overall is spent using DR KPIs. The good news is that Brand spending is a leading indicator of economic health - as soon as Brand spending increases, you should expect to see the economy follow.
|
|
Published:
2/13/2009
By the way - I made a bit of a mistake above when I said SnapTell didn't have any demos on their site. For some reason I missed this.
If you take a picture of the cover of any DVD, CD, Book or Video Game, send it to fun@snaptell.com, you will receive a text message almost instantly with a link to a mobile web page with information about that product. That's pretty cool - and it worked flawlessly - and believe me - I tried some obscure things.
|
