Lost in the Cookie Debate--What About Data Harvesting?

AMID THE HUBBUB ABOUT COOKIES and their uses over the summer, one element of the debate seems to have been lost in the mix. Ironically, it is among the more prevalent uses of cookies, and also among the more potentially dangerous uses for publishers.

I'm talking about the instantaneous data harvesting that occurs routinely in the online ad buying/delivery process. Here are two types that have made a few companies and individuals extremely wealthy, while remaining predominantly under the radar of this debate.

Ad Broker Networks--these companies (like Advertising.com, Fastclick, Undertone Networks, and others) place cookies on consumer browsers as part of their ad buys. They typically capture and store information about where the person was surfing when the ad was delivered (such as the technology section of USAToday.com), what they just surfed before reaching that site (the referrer's Web site), who the advertiser is and what type of business/or offer is involved (i.e., Sprint, mobile telecom offer).

Additionally, this cookie obtains a great deal of technical information (the type of browser, browser plug-ins, machine brand, time of day, connection bandwidth, and IP-based geo-location), and finally, whether or not the surfer clicked on the ad.

These ad brokers capture this information to optimize the space that they bought (their business is to buy low and sell high) with CPA offers. However, they typically keep this information (both on the cookie and in a database, most likely), and re-use it whenever they see the user again. For example, they may routinely make large, multi-billion monthly page view buys on MSN or Yahoo! at a very, very low CPM--maybe for less than 40 cents. If they know that a user visits tech content at a major publisher site, is located in New York City, and has responded favorably to a mobile telecom offer recently, they could deliver that person a very high-value Verizon Wireless ad with an effective CPM of $10.

Essentially, the broker is able to leverage cookie-based data that they captured from one publisher on an earlier buy (which paid for itself) into a massive arbitrage on a subsequent buy solely because they were able to capture and re-use cookie-based data.

This is one of the major ways that ad brokers make money, of course. Rarely do publishers ever ask or require that these networks disclose which cookies they will be setting when they make buys. Nor do they require knowing what data the brokers will be capturing, or what they will do with it.

In fact, many times, the broker that buys the remnant space will not be the one that actually fills it. For example, a company like Drive Performance Media (DrivePM) may buy space from a publisher. They may have their ad server called first. If they don't see a cookie that they recognize, they may "hand-off" the space to Blue Lithium, who may do the same thing. They will look for one of their cookies. If they don't see one, they may hand it off to Right Media, who may resell the space again.

This kind of tactical data harvesting is almost always undisclosed to the publisher, and is certainly unknown to the consumer. Is it bad? Well--for publishers, it may be. But many publishers are happy to part with their remnant inventory under these terms. Is it bad for consumers? As eCommerce grows and the Web marketing model expands, that would be an even more difficult case to make.

Should it be more transparent? That's what I'm hoping to hear from you readers on. I think it needs to become more transparent--and soon. But I'm assuming many of you will disagree with me on that point.

The second type of Data Harvesting I'd like to examine here is done by Third-Party Ad Servers. Some third-party ad servers set a unique cookie in their own domain on all browsers of all campaigns that they see. This way, they are able to see this cookie in the future when they see the same user again, assuming that the user has kept the cookie.

Some third-party ad servers routinely keep technical information about the user with this cookie (browser type, etc.). Some third-party ad server contracts include as a standard default clause the right to keep and re-use certain anonymous data about the campaign. For example, they may keep information about what category of ad was displayed and how the user responded. They may keep information about what type of Web site or section or page the user was surfing or referred from when the ad was delivered. They might keep information of the frequency of visits or how recently they have seen the user. Some of these companies then re-use that data for campaign optimization as part of proprietary "optimization" that they charge advertisers for. They may also use this data much like the ad broker networks, and buy remnant inventory to then optimize it against CPA campaigns.

Atlas DMT, for example, may make purchases like those aforementioned on MSN to optimize their inventory, or they may just offer the optimization as a service. Essentially, they are able to convert the consumer data that they captured from publishers and from advertisers or other agencies that use their services into media arbitrage.

As consumers, are you aware that this is going on in your hard drive and vis 'a vis the sites you visit? More to the point, if you're a Web publisher, did you have any idea where your user data has been going and who else has been monetizing your users?

For years, I've been saying that the cookie/privacy issue is far more of a business matter than a consumer privacy matter, and I think that is truer still today than ever before.

But until Web publishers get the same kind of transparency from their affiliates that consumers are demanding from the publishers themselves, there will be far less industry awareness than is needed in this debate.

Tech Q&A: What are cookies, and are they dangerous?

By Anick Jesdanun, The Associated Press

Q. I've been hearing a lot lately about "cookies" on my computer. What are they, and are they really dangerous?

A. Cookies have been getting a bad reputation lately because some are linked to spyware and adware, programs that often sneak onto your computers. Many anti-spyware programs identify cookies as threats because Web sites can use them to target ads based on your surfing habits.

But cookies are much more than that.

Fundamentally, cookies are small data files your Web browser uses to help sites remember who you are. They were invented at Netscape Communications Corp. a decade ago and incorporated into the then-reigning Netscape Navigator browsers to make Web surfing more efficient. Other browsers soon adopted cookies, too.

Without cookies, each interaction with a Web site would be treated as a new visit. The site would have no way of knowing that the page you just called up and the one you called up five minutes ago were summoned by the same person.

That's often not a problem for basic Web pages.

But for commercial Web sites, it can be.

Major Publisher Discusses Targeting

USA Today's Vice President of Sales Lorraine Ross discusses behavioral targeting with iMedia's Neil Perry.

Coming soon from iMedia Communications is a new report, "Demystifying, Defining and Profiting from Behavioral Targeting." For the report, iMedia Vice President Neil Perry interviewed a wide range of industry experts, including USA Today's Vice President of Sales Lorraine Ross. While some of her most important insights appear in the soon-to-be-released report, here we present an extended excerpt from the interview.

Neil Perry: Who is your provider for behavioral targeting, and how long have they been doing it?

Lorraine Ross: The provider is TACODA, and we have been doing it since … I think we launched it September 2003.

Perry: What is the benefit to USA Today of behavioral targeting?

Ross: The benefit to us is it helps us optimize our inventory by putting the focus on the characteristics of the audience, regardless of where they are on the site. You create value by identifying the audience, not just the content they are reading. And, so, you multiply the value. It is not like you are abandoning the value of the contextual relevance; but you are able to say, "Oh, and in addition, we know this about this person."

The less anonymous our reader is to the marketers, the more valuable he or she becomes.

So, it optimizes the inventory. It also helps to drive rates, because you are, to a certain extent, profiling. You are able to profile me, the reader … you are able to get more for exposing the marketer to that reader, that audience member.

Perry: What about for the marketer? How do you serve it up to the advertiser? What is the benefit to them?

Ross: Well, better targeting. I think that there is … the ability to provide more and more targeting against the individual, not necessarily in a [direct response way], you know clickthrough, or take-action kind of way; but, just in terms of being able to identify demographics, behavior (meaning sort of intention, or mindset), geography -- all of these things are incredibly valuable to a marketer, because they are able to speak directly either their best customers, or best prospects.

Perry: All right, now how satisfied would you say you are with the success of your behavioral targeting program?

Ross: I am very satisfied. It is still in its infancy, though. I think that behavioral targeting still has a long way to go as we continue to, first of all, grow our audience and collect and aggregate information about our readers. You know, for example, we have a mortgage calculator on our site. We have looked at creating a behavioral segment around it, because certainly that kind of purchase intent is ... to be able to identify that is phenomenal -- it is phenomenally valuable. But, we don't have enough people at any one time really using the mortgage calculator on our new site, (because we are not a financial site specifically), for it to really make sense for us now.

But, as our audience, as our site continues to grow, or we link into other sites, then that potential is really there to create those segments. So, we are very satisfied with the contribution that behavioral targeting has made so far. But, I think we continue to look out to say, "All right, what else can we do with identifying our readers? How else can we do that? And, how can we even cross path the characteristics? Can we layer things on top of one another to make it that much more specific?"

Perry: Have the marketers who are utilizing it gotten a little skittish of late, with all of the HR29 [anti-spyware legislation] related stuff that is going on out there, and all of the focus on cookies? Or, are they still pretty actively engaged in behavioral targeting?

Ross: They seem to be actively engaged. They care about their success metrics. And, from what I can tell, there is full movement away from rigid adherence to success metrics, such as, "Well, we need people. We don't care how you do it, Mr. Publisher. But, we need people to visit our site and go through three page views, or we don't consider it to be a success." So, they are really, really still very, very busy evaluating on the back end. If we could do it by only serving ads on Tuesdays, and that got them their success metrics, that would be fine with them. It is really the publisher that is saying, "I think I know what would work for you." We are trying to answer the marketer's problem. But, the marketers … I don't even think they care how the problem gets solved. It is really up to the publisher. But, they recognize logically behavioral targeting makes sense as a way to attack that problem.

Perry: Approximately what percentage of your campaigns are behavioral targeted, or include a behavioral targeting piece? And, where do you think it is going to go next year?

Ross: I think there is never a campaign that is solely behavioral targeting. That is largely because we are trying to optimize a client's campaign. They come to us, and we say, "Okay, a little of this, and a little of this, and a little of this." A little contextual, a little behavioral, a little demographic targeting. And, then that gives the publisher the flexibility to, when the marketer comes back and says "This part is really working, but this part is not," you are able to really hold them to the buy. But you know, you start out with three different options, three different choices. So, there is an inherent expectation that the client is going to come back to want to optimize.

But, to answer your question, and what I will do is I will go back to finance and ask for the percentages, because the [estimated] amount of … the number of proposals, or IOs [insertion orders], that actually contain a behavioral targeted component is about 50 percent. But, the amount from behavioral targeting very, very specifically and narrowly stated is maybe, you know, 10 to 15 percent.

Neil Perry: What are your thoughts, and your company's thoughts, about all of the broo ha ha over cookies, and the deletion of cookies, et cetera.?

Lorraine Ross: You know, we have read the Jupiter survey, and I keep up on all of my newsletters, and … I think they are right. I think that there are an awful lot of people who are deleting their cookies, even inadvertently by using AdAware or Spybot. Keeping your PC clean of viruses has become, I think, critical to consumers. And, if by using those products you are wiping out cookies, then, yeah, I think cookie counts are incredibly unstable. So, I think that there is a real issue there. I do. And, I think that it is really problematic. It sets us back, because it prevents us from really understanding a lot of things -- the total size of your audience, frequency -- all of these things are compromised when users are aggressively deleting their cookies. I mean, you run the risk of asking them again for registration information, and it is just … you know, you really get worried about that.

Perry: Yeah. The consumers are getting, I think, much more reluctant to offer up that information, just because those one or two times you get taken advantage of it is just a nightmare.

Ross: Well, and then you … especially when the registration process is a little bit, somewhat onerous. Then, your cookie gets wiped out because you are using an AdAware program because you are trying to eliminate viruses. And, then you go back to your favorite site, and they want the same information all over again. And so it is a slippery slope.

Perry: Yeah, like when you ask just to get a copy of an article, and six screens later you are still filling out some forms, and you start realizing …

Ross: Yeah, I think that that is really a problem. That is why we have made a conscious decision not to go … and here is where the difference is: I think the companies that are contact companies exclusively are going to have a really tough time maintaining a customer information database and asking people for information. I think companies that are utility based, and I mean media companies like Yahoo! … well, you are really going there to use a tool, and the contact is really a nice side business for them. But, you are really going there because that is where your email is. Or, you are going to, you know, you are going to use a tool. Then, I think that they have a better chance of … you would be more likely, more inclined, to go ahead and put that information back in again.

I also think that … (I forgot to mention this earlier) one thing we have done is we have undertaken what we call the "perceptive cookie project" here, where we try to … you probably can only host, I guess it is anywhere up to 20 cookies, at any one time. And, so the cookies will get bumped off if new cookies are being introduced. You just can't host that many. So, we also have tried to do that where certain cookies will be made persistent. It is almost like you put a timestamp on the really important ones. And, you push those to the front of the line. It doesn't eliminate the problem of cookie deletion. But, at the very least, you can take the most important cookies to you … or, the ones that are, you know, site designated, and then make sure that they are the freshest cookies so that they never expire and get bumped off of a browser.

Perry: What advice do you have to either marketers or advertisers that are considering behavioral targeting?

Ross: Ummm … advice. What really works best is when marketers come to us with broadly stated issues, and they help us to solve the problem with them. Because, most of the time, we don't have a canned segment created for them. But, if we apply our brains and some technology we can come up with a customized solution. I think increasingly sites will be required to really come up with more customized approaches to a marketer's problems. And, behavioral, I would include in that. Some of these solutions are only going to work if the site is gigantic -- like one of the portals. But, that is the advice I would give: when you receive directions from clients that are really very tactical … and I get really worried about behavioral standards, because I think, well, one size is not going to fit all. I mean, we might be able to solve your problem without necessarily adhering to an arbitrary standard that you are setting for the behavioral segment. Let us try some different things with you, to try to solve that problem.

Perry: Thoughts on the future?

Ross: Where it is going to go to next is, I think, really interesting, because publishers are increasingly … technology is being developed, software is being developed, that allows publishers to database all of their information and metatag all of their information. And, increasingly, publishers are moving in this direction. We will be able to take advantage of some of the … not so much Google's business model, if you will, by observing the behaviors, on a nuance level on our site. So, for example, publishers are moving towards the dynamic display of information.

So, you come to my site, and you are looking for information about a certain band, say, in the music section. And, the publisher at this point will be able to, very soon, offer you up a page -- and, it really doesn't even exist on our site, but because you requested information about a specific band, we are able to dynamically compose a page of relevant articles, photos, video … you know, links right in front of you.

And, I think where behavioral targeting goes after that is that now I know that you have this interest. And, you have used my site in a way that maybe you formerly would have used a search engine … to be, just to be entertained and get your information. And, that is going to provide publishers with much more insight into their behavior and purchase intent, rather than just, "Well, what did you tell me in your ZAG [ZIP Code, Age, Gender] registration form? And, have you gone to the travel section, you know, a couple of times in the last 30 days?" So, I think publishers, and the publishing model, are really shifting in an interesting way that will allow us to -- without invading anybody's privacy -- be able to really infer a lot more about our readers and be able to offer that kind of relevant advertising that everyone says that they are interested in. So, that is going to change, and I look forward to that, because it is better targeting, better targeting, better targeting!

Perry: Yeah, and really, that is the strength of our business. And, we are really starting to assert our muscle, right now, in that area. So, it is great.

Ross: Yeah, I mean, publishing is changing, and that is going to have a direct impact on our ability to understand readers and serve them better advertising.

Perry: Yeah. Great. Well, listen, I really enjoyed this conversation, and I appreciate you taking the time.

Ross: All right, take care.