nugg.ad @ Interact Congress

This year´s Interact took place in Berlin and a case study from Ekstra Bladet was presented there, showing the effectiveness of nugg.ad´s predictive behavioral targeting. Please find the presentation in the following.

| View |

iMedia Summit in London

Well, to be honest with you, the iMedia Agency Summit did not really take place in Central London but rather St. Albans, pretty much on the outskirts of London in a very nice and very British countryside (with a golf course in the neighbourhood). Frank, Jesh and myself had a lovely stroll (in the mud;-) enjoying the fresh air (of sheep and liquid manure;-).

The event itself Read the rest of this entry »

Google updates it’s data sharing policy: Analytics Data will enhance Google-Ads

Today is saw this announcement in my Google Analytics account:

Google Analytics Data Sharing SettingsIn order to improve your experience with Google products, Google Analytics is updating its data sharing policy. Read the rest of this entry »

Data retention and behavioral targeting: why targeting via “service providers” is (can be) the best option*

Friday was a black day for civil rights and informational self-determination. I grew up in Bavaria, but after I moved away I had hoped that the whole business of computer surveillance, dragnet controls (“identity checks where there are no grounds for suspicion”) and suchlike would belong in the past. Now these measures are being introduced throughout the whole country, on my beloved Internet. To connect this with September 11 is laughable – anyone who has studied the material knows that. So should we read the newspapers more again, and watch more TV? Because no IP addresses can be tracked there?Hopefully not (although reading the papers more often wouldn’t do any harm – but that’s another story).In addition, there is an increasing number of initiatives dedicated to the craze for storing data on the Internet which want to fight against it becoming simply a matter of course to collect and link data and to use the resultant personal profiles for whatever arbitrary purpose (see for example: www.wirspeichernnicht.de). At an FTC hearing it was even suggested that a do-not-track lock list should be set up.I think that everything possible should be done to keep the Internet as a place where people can distribute information more quickly, with fewer filters and with far lower entry thresholds for authors and readers. In other words, as a sensible and intelligent new form of reading newspapers (and producing them). But also with free offerings and simple sources for financing this…That’s easily said, of course: of course I’m writing in the nugg.ad blog and I work for a company that promotes itself with the claim to make advertising more intelligent and better tailored to the users’ interests. I’m also convinced that there’s no way this trend in online advertising can be stopped, and it must therefore above all be structured. In addition, it’s an incredibly exciting project to finally give online advertising the features that are possible online and which put online ahead of all other media. And that is also to do with tracking…The point is: no personal profiles need to be created to do this! I’d even go so far as to say that no one in the advertising trade actually has any real interest in Schäuble-type personal profiles which can be related to individuals. A rough “good guess” is perfectly adequate. A segment that is served efficiently, a halfway passable estimate of what might interest someone (we’re talking about Flora margarine vs. Stork margarine and the like here, not about political leanings or sexual preference). What’s not clear to a lot of people is that advertising for micro-profiles or even individual persons is simply unaffordable and consequently not wanted. Even the popes of CRM Peppers & Rogers soon distanced themselves from the strategy of 1:1 marketing and have introduced “mass customization” as their guiding principle.This should be grasped as an opportunity for behavioral targeting. In other words, not only is it prohibited to create individual personal profiles without permission; it’s just not, in fact, necessary! So it shouldn’t be done; on the contrary, every effort should be made to avoid sailing in the wake of these unspeakable initiatives (data retention, etc.).So what should a company do when it wants to introduce “clean behavioral targeting”?

1. Ensure that no IP addresses are collected or stored, preferably by using a strong 3rd-party anonymizer and ensuring corresponding contractual agreements, because this is the only way to guarantee that the protection cannot be bypassed. That’s the ultimate measure to avoid becoming an involuntary accomplice of Schäuble & Co!
2. Don’t work with finely granulated click profiles (“User A clicked on football 10 times, NRW regional news 4 times and adult entertainment once”). A statistically concealed profile is quite sufficient (“User A has 0.8 interest in football, 0.3 in adult entertainment and 0.4 in news”).
3. Don’t do it yourself! It’s a common mistake: implementing behavioral targeting in your own company, in particular when you also have access to other data of the user (because you are the ISP, for example) is in no way more reliable than putting it in the hands of a third party. The danger lies in the inadmissible linking of data, and this is far more transparent and must be regulated by clear agreements when targeting is implemented by a specialized service-provider. Data privacy experts speak of the separation of information powers here.
4. Offer an opt-out option and inform the users clearly about how the data are used. The request to the FTC to introduce the do-not-track list is justified by the fact that the industry has so far not been able to create easily accessible opt-out options voluntarily! (I think the authors know the global opt-out site of NAI – I think, from their point of view, that this option should be given more prominence and offered directly on all content pages…)

By the way, the most effective measure to gain user acceptance is unfortunately not on the list yet, but I’m sure that will change sometime or other: a sensible trade-off! When users see that tracking (which is performed according to clean and transparent principles, see above) provides them with advertising which is really more relevant (and perhaps a smaller volume of advertising) and that a gem of information can perhaps be found in it sometimes, they will have absolutely no problem with it! Users want to get something in return for the data they provide. Currently they get too little, even though it can be proved that behavioral targeting functions excellently – for the industry, that is…Until this is achieved, the same applies for online advertising as ever: behavioral targeting will help to maintain the growth of online advertising at a high level. Such technologies will also attract new budgets to the online sector from the TV and print sectors (keyword: Lätta). Nearly all the exciting, free offerings on the Internet are financed with this money – and currently really only with this money! Regardless of whether a good content site is concerned, such as Spiegel-Online, Google Maps or free email. Or some fancy Web2.0 game site. Without online advertising, none of them would exist. Consequently everyone (publishers, users, advertisers) should be interested in finding the right way to deal with the topic. Closing your eyes doesn’t help; rather, keeping your eyes open, but without hysteria, by implementing sensible measures. And the industry must simply do everything possible to build up trust honestly.So: we do perform tracking, but what we track does not result in a personal profile. Tracking is therefore no more dramatic than the fact that your way through the supermarket is recorded by market researchers and is used to optimize the shelves. They’ve been doing that for years, by the way…*PS: Naturally I’m not trying to say that in-house targeting per se is not OK. The principles outlined above can also be used for that – even if by definition introducing the separation of information powers requires another party to be included…

Facebook targeting: but how?

I don’t want to be a spoilsport, and at the moment it’s about as dodgy to criticize anything about Facebook as it is not to buy an iPhone.But I have to say that all the hype about the wonderful Facebook targeting (and also about what Myspace has launched) puzzles me. Even more so now that I’ve taken a closer look at it (yes, I have a Facebook account!).I found it fundamentally rather questionable that sensible targeting for online advertising should be based on a self-maintained community profile. Shit, now even the holy Web 2.0 principles are coming under attack… Surely we don’t want to call into question the user-driven content, tut, tut, tut. But nevertheless, profiles created in Myspace and Facebook are simply a long way from being taken seriously. After all, they’re not a virtual registration office. People present themselves there and turn themselves into what they’d like to be, and so on.That’s totally OK, and it’s a no-brainer. But why that should suddenly be so revolutionary for online advertising really puzzles me. And it’s all the more puzzling when you examine exactly what a Facebook profile can actually contain, provided someone has really taken the time to complete it and the details are half-way correct.Let’s just imagine we’re a brand manager of an online advertising budget and we are confronted with this online advertising revolution. What criteria can we now choose for out advertising?

1. Sociodemography

OK. Age, sex, hometown, country, political views and religious views. All regional information is old hat, any old ad server can do that, and it doesn’t work anyway.Age and sex are fine, of course – no question. If these details are correct (??) they provide important information for online advertisers! Though some email providers and ISPs have got this information – if they really dare use it for advertising…But the revolutionary potential is still barely above zero.

2. Relationship status

Are we hetero or gay? Married or “complicated”? Are we looking for sex (openly or secretly)?This is irrelevant for advertising – except for online dating sites, perhaps.And with dating sites you can already see another problem: how would it be for users if they entered “Looking for Dating” here and then the banner for the Parship dating site would always appear at the top of the screen? I think people would find that most unpleasant…

3. Interests

Now things are getting interesting. Personal interests and preferences. With a free entry mask. Hmm….Let’s be honest here: how much information will we find here that is relevant to advertising? Not even taking into account different spellings and so on. But what advertiser is going to be interested in my favorite films or books or what I consider the coolest quotes? I believe we’ll soon witness a few wonderful cases concerning music downloads or DVDs (my tip: Hitflip). Those will work too, if reach is not taken into account. So we have the simple phenomenon that users (naturally not all of them) will specify something here, but these details will be far from complete and certainly not up to date. If I think the new Babyshambles CD is great, I’m going to rush to update my Facebook interests profile before I do anything else. And I’m not going to enter all the songs in my iTunes library there, no, no, no…The result is simple and dramatic: Facebook targeting has no reach! Even if a few users mention Babyshambles, Facebook will still only contain a fraction of the Babyshambles’ fans. Even Facebook comes quickly into the region of just a few thousand profiles. And these then have to be active for the campaign. But advertising aims at bulk, and the TKPs that call Facebook will only work with bulk. Ergo..…But above all: from the point of view of advertising, how much is missing here in terms of interests? What about consumer goods, electronics, insurance, financial services, automobiles, DSL preferences, hair conditioners and the wonderful “sweet and savory snacks” (AGOF)? None of them are there. And even if an entry form were available for these products, of course nobody would complete it – it would probably drive the users away. Or they would bombard it with spam, like the wonderful interests questionnaires that gmx regularly sends its users, for example.What else is there? University, current employer, previous employer, job description, etc.

Conclusion: Facebook will permit targeting with respect to age and sex. That’s it. And good advertising opportunities for dating sites and music labels perhaps – but with major reach problems and tricky questions relating to data privacy.

For advertising, it’s the other way round: forget the profiles, use them for profilingI have no doubt that Facebook and other social networks work…as social networks! In other words, we will actually find a great deal of relevant information on people there, often private information and not infrequently also quite special things which people describe well. That’s the whole point of these networks – that people use them to describe how unique they are (definitely with a mixture of fact and fiction). And then that they find other people in this way (because they want to do this, the quality of the profiles will always be totally OK). But the profiles will never be of the type that the advertising industry would want. However, with statistical methods, a few additional tricks, and intelligent handling of the topic of data privacy, you can make something out of the profiles that will be relevant for advertising.That is, when you take the people’s digital DNA as a basis for profiling procedures and don’t attempt to control advertising directly on the basis of the profile information! For example, by questioning a few Facebook users about the products they are interested in. And definitely not entering the answers in the profile, but keeping them strictly separate – and also communicating this fact. That’s the only way to get honest answers (as it simply makes no sense to give fake answers). Then take these answers + the digital DNA as the basis for profile assessments.

That will work; I’m sure of that.

Is Predictive Behavioral Targeting only something for geeks, nerds and freaks?

No. But admittedly the concept of Predictive Behavioral Targeting is not always easy to convey. It’s not just that we are replacing advertising for environments with advertising for people on the basis of their past behavior. No, we are also working with the results of live market research and complex algorithms. In some meetings with customers I have therefore felt almost physically how difficult it is to succeed in the face of what has been learnt over decades of media planning and marketing. Naturally terms such as Rich Media and Skyscraper are sexier than an SVM analog for the Bayes hypothesis (cf. Stephan’s posting). I am therefore all the more pleased that despite, or perhaps precisely because of, our geekiness the generic term Predictive Behavioral Targeting which we coined is becoming increasingly accepted and understood. While a Google search for the term in January only returned 20 hits – all of which, moreover, referred to the nugg.ad website – the figure now is 3,000 and still further increasing. The highly-recommendable eMarketer report by David Hallerman titled “Behavioral Targeting: Advertising Gets Personal,” which contains predictions under Targeting 2.0, has certainly played its part here: “This track for behavioral targeting looks to develop technologies that predict – rather then harvest – customer attributes that are based on page views, searches and other online behaviors.” Here are a few extracts. To make it easier to read, just click on the graphic: