The Fallacy & Conundrum of User Influenced Ad Models

As the online display ad business continues to focus more and more on the idea of user targeting (the idea of targeting the user instead of targeting the site or the context of the page), there is a growing interest and some potential concern around how we’re going to target users.  With some of the more extreme ad models now scaring the bejeebus out of users (see Phorm and Nebuads) and growing concerns about the hegemony of companies like Google and the data it’s collecting about users (see the lack of privacy built into Google’s new browser), some companies are going creating ad networks and systems based on the ideas of users themselves giving explicit feedback about ads they like and don’t like.  Let’s just say I think that any ad model that relies on users giving feedback is a disaster and doomed from the start.

Filed Under – Doomed to Epic Fail

On paper, the idea of allowing users to give direct affirmative consent and feedback about ads they like, things they want, their interests sounds very democratic and utopian.  Power to the People! and all that.  The problem is that any model that relies on users doing anything other then what they really want to do flies in the face of what people actually do online.   Despite the web being about interaction and participation, almost 99% of what people do online is read, not participate.  At any UGC website – only a small portion of the audience actually ever uploads anything.  At YouTube for instance, I’ve heard reports that despite the 100 million users it sees everymonth less then 600,000 users ever upload and share anything publicly.  600,000 might sound like a lot but it’s less then 1% of their user base.  At FoggyGames.com, the casual games website I own, the percent of users who’ve ever rated a game – which only requires a nano second of effort to click on the classic rating star – is less then 1% as well.  Again and again you see participation rates in that range.  So now these new ad models expect the vast majority of users to actually rate each and every ad – even when they have shown again and again they won’t even participate in sites and actions where they actually want to participate – I don’t think so.

Sampling Don’t Work

OK then what about the idea that you don’t need every user to participate that just getting that sample to tell you about what ads they like or don’t like.  Unfortunately the idea of a sample defeats the whole principle of user targeting.  The basis of user targeting is targeting a specific’s user definite demographics, intents or interests.  And again and again – we’ve seen that sampling doesn’t work since it is the way that most site level targeting works today.  Sites sell ads based on samples of their user base – their user base is 60/40 Female/Male and thus they sell a disportionate number of ads targeting females (it’s largely what Glam Media does across multiple websites – not very sophisticated technically speaking).  Thus lots and lots of men get poorly targeted ads just because they go certains websites in this example.  The whole idea of user targeting is to solve that problem – show ads to men with ads for men in that scenario instead of generic ads based on a sample.  So without finite, user level data no user targeting scheme can work.  So in a web where you’ll likely get more then 1% of users to participate, models that requires something much greater then 10% and more like 25% of users to participate seems like a folly even at the start.

Conundrums and Contradictions

The conundrum and contradiction with user targeting is that users say they don’t like being tracked.  Yet what they won’t do is explicitly tell the advertiser or advertising provider what they want.  And yet, again and again users say they want ads targeting to them individually as way to increase the quality and relevancy of ads.  I’ve done enough research to know users like relevant ads – they actually stop them ads and start calling them information (ads are pejorative term meaning noisy, irrelevant, uninteresing, annoying marketing messages).  Tying a user’s information and interests from where users express them willingly (communities, social networks, etc.) to where they get exposed to ads is a way to solve that.  Doing so in a conscientious and respectful manner is critical for all players in the market (trust me I speak from experience – one bad actor can sink a market) if we want to solve the user targeting and participation problem.  And the folly is we’ll be able to avoid that get by getting the all the users online to vote on each and every ad – I’ll hopefully save everyone some time and investors money ‘cuz it ain’t going to work.

This entry was posted in work and tagged , , , . Bookmark the permalink.

3 Responses to The Fallacy & Conundrum of User Influenced Ad Models

  1. Well said. There’s a term for the imbalance of participation on the Web — “participation inequality”, and I wrote about it here at http://kickstand.typepad.com/metamuse/2006/10/of_users_who_ac.html. Jakob Nielsen did some studies on it.

  2. Neal Richter says:

    Nice post. I disagree on “sampling won’t work” in general. How and what you sample and apply the rules/knowledge learned to makes or breaks sampling.

    Your example is a bit of a straw man as doing a 60/40 split on men versus women ads to random visitors is horribly stupid… when your rule learning method tells you to flip a coin then it’s FAIL.

    Here’s an example that works: Learning that there is a correlation to men buying beer and diapers together. One only needs to learn that correlation over a representative sample of transactions.. consuming all of the data is unnecessary unless you are looking for the low frequency correlations.
    If those infrequent correlations matter, then you can still sample if you do it smartly.

    Second example: learning that your male viewers visit different pages than the women or enter via a different route, then you’d be able to target ads to them correctly. If page-type X – then assume men and show mens ads. You can learn that rule with sampling and smallish tests if done correctly.

    The devil’s in the details.

  3. I TOTALLY agree that User Generated Content is a myth – indeed, 99% read and only 1% contributes.

    However, I like your conclusion that users start calling ads information as soon as it becomes relevant to them.

    Explicitly asking someone whether or not they like the ad, defeats the purpose of the ad in and of itself. Rather, advertisers should be asking whether users like their products and what they could do to improve it.

Leave a Reply to Richard van den Boogaard Cancel reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>