W3C home > Mailing lists > Public > public-tracking@w3.org > July 2012

Re: Frequency Capping

From: Chris Mejia <chris.mejia@iab.net>
Date: Wed, 11 Jul 2012 23:49:44 +0000
To: Jonathan Mayer <jmayer@stanford.edu>, Tamir Israel <tisrael@cippic.ca>
CC: "Grimmelmann, James" <James.Grimmelmann@nyls.edu>, W3C DNT Working Group Mailing List <public-tracking@w3.org>, Mike Zaneis <mike@iab.net>, Brendan Riordan-Butterworth <Brendan@iab.net>
Message-ID: <CC2380DC.1FCC5%chris.mejia@iab.net>
Jonathan,

What exactly did I misunderstand?  I must have missed your pure 'software engineering' argument— perhaps because it was wrapped up in misguided conceptions about why f-capping is conducted in the first place.  Still, I fail to see a pure software engineering discussion in your post critiquing Brian O'Kelley's post.  Also, I must have missed where you stated the actual real-world concern around f-capping that you are trying to solve for?  Please elaborate—I'd love nothing more than to be enlightened on these fronts.  As you know, I don't much appreciate solving for problems that don't actually exist—I prefer to solve for real-world problems, not edge cases.  And through this entire thread, I have not seen one person point to the actual real-world problem with f-capping that needs to be solved for.  Actually, through productive back and forth dialog, some have come to understand and appreciate the practice and its need.

Also, based on your statement below, I wonder if you'd support the decoupling of the technical spec from the compliance document?  I agree, it would be great to have a pure technical spec that's not based on policy concerns at all.  You seem to agree?

Finally, your obvious lack of real-world knowledge and appreciation for the industry you propose to regulate greatly concerns me, as you have selfishly promoted yourself in the press as an "expert".  So, if you are not on board with learning about the space and our justifications for currently accepted and standard practices (so basic as frequency capping that represents a clear win-win for consumers and industry), how do you propose that your so-called 'software engineering discussions' will be of benefit to the discussion?  Software engineering discussions are only productive when they are relevant.  I'm part of the forum, learning about your privacy concerns—please pay us all the respect of at least TRYING to understand the business you are proposing to regulate.  Bring real problems; justify them; cite examples; support your case.

For everyone just seeing this who might be trying to understand where I am coming from, Brian O'Kelley's post to Ed Felton's blog on frequency capping is available here: http://techatftc.wordpress.com/2012/07/03/privacy-by-design-frequency-capping/.  In his reply to Prof. Felton's post, not once did Brian cite "revenue" as the justification for f-capping (in fact the term "revenue" was never used in his description of f-capping).  Yet Jonathan's post here on the W3C forum "recounting" (poorly paraphrasing) Brian's description of f-capping talks only to so-called "revenue" justifications— sorry, that's just not it Jonathan— not only wasn't it in there, it wasn't in there because it's not right.  I've pasted Jonathan's post below, so we can ensure we are all on the same page.


Chris Mejia | Digital Supply Chain Solutions | Ad Technology Group | Interactive Advertising Bureau – IAB


I'd sure like to hear more from advertising industry participants about how frequency capping integrates into advertisement selection.  The AppNexus approach, if I read correctly, goes roughly as follows:

1) Begin with the set of all campaigns.

2) Filter by targeting criteria.

3) Filter by frequency capping.

4) Assign an expected revenue to each campaign.

5) Select the campaign with greatest expected revenue.

The approach includes testing the frequency cap of every campaign that matches targeting criteria.  What about, instead, only testing the cap for a subset of those campaigns:

1) Begin with the set of all campaigns.

2) Filter by targeting criteria.

3) Assign an expected revenue to each campaign.

4) Select the n campaigns with greatest expected revenue.

5) Filter by frequency capping.

6) Select the campaign with greatest expected revenue.

Some relevant empirical questions include: How often are the highest revenue campaigns frequency capped?  How well can an ad company predict which high-revenue campaigns will and won't be frequency capped?

Jonathan
From: Jonathan Mayer <jmayer@stanford.edu<mailto:jmayer@stanford.edu>>
Date: Wed, 11 Jul 2012 15:26:33 -0700
To: Tamir Israel <tisrael@cippic.ca<mailto:tisrael@cippic.ca>>
Cc: Chris Mejia - IAB <chris.mejia@iab.net<mailto:chris.mejia@iab.net>>, <Grimmelmann>, James <James.Grimmelmann@nyls.edu<mailto:James.Grimmelmann@nyls.edu>>, W3C DNT Working Group Mailing List <public-tracking@w3.org<mailto:public-tracking@w3.org>>, Mike Zaneis - IAB <mike@iab.net<mailto:mike@iab.net>>, Brendan Riordan-Butterworth - IAB <brendan@iab.net<mailto:brendan@iab.net>>
Subject: Re: Frequency Capping

I believe Chris Mejia entirely misunderstood my note yesterday, and the thread has since careened into a conversation about the business relationships and economic value associated with frequency capping.  While those topics appear to have been educational for some participants—please continue!—that was not at all my aim.

I want to have a software engineering discussion about how frequency capping is currently implemented, and how it might be implemented in ways that better protect consumer privacy.  The CEO of AppNexus was kind enough to give details of his company's implementation which, as he explained, does not neatly integrate into a privacy-preserving approach.  Leonid Litvin from PulsePoint suggested that the algorithm I proposed—which is compatible with a privacy-preserving approach—might work.  Let's pick up from there.

Jonathan

On Wednesday, July 11, 2012 at 2:29 PM, Tamir Israel wrote:

OK Chris, I agree. I think my point was that DNT-1 is less a rejection
of the value exchange than, say, AdBlock or a similar plugin.

I understand that targeted impressions are worth more and I've heard
they generate more click-through.

I simply meant to say that DNT-1 a.) still allows impressions; and b.)
still allows contextual targeting (by site, etc.), so its value is not '0'.

Two quick side notes:
I am not remotely convinced this spec is going to lead to ubiquitous
DNT-1, and I don't think this working group is currently considering
anything that might make this the case; and

Also, I am no longer saying there is no value to F-capping for DNT-1s.
It makes sense to me that at least some types of advertisers would want
to just reach 'everyone' so would purchase, say, 10 million impressions
hoping to reach 5-10 million people (whether targeted or not). On this
scale, there is a definite risk of a DNT-1 user seeing the same
advertisement more than once, and also there is a benefit to maximizing
the ad campaign's reach, as desired, so some form of frequency capping
would seem to have value.

On 7/11/2012 5:08 PM, Chris Mejia wrote:
Thanks Tamir. I stand corrected--consumers who elect to express DNT:1 MAY
not have completely opted out of the value exchange, you're right.
However, their relative value to the value exchange certainly goes down.
To further explain, when users see un-targeted (randomly placed) ads that
are not based on their general interests, they are likely to ignore those
ads. In ignoring those misplaced ads, it's a double-whammy on industry:
we pay to serve ads that the consumer will never engage with, nor buy
their products/services. Obviously this decreases the relative value of
that consumer engagement and lowers the overall revenue the publisher may
charge an advertiser in connection with the serving of the advertiser's
ads to that non-targeted consumer. In this case, f-capping would be even
more important from a cost-savings perspective; the more non-relevant ads
I serve a consumer, the more cost associated-- f-capping limits delivery
and thus limits costs. Also, it's probably not a stretch to assume that
many advertisers may not want to serve their ads at all to consumers who
are expressing DNT:1. Enter the digital divide once again: anti-targeting
may lead to a situation where the only ads being served to 'lower-value'
DNT:1 users are the ones everyone would rather avoid (annoying content ads
that are served only on a CPA basis). Premium content ads are generally
very expensive to produce and serve (premium rich media ads cost more to
serve), so my educated guess is that advertisers wont want to take a
chance on where they will spend money serving these ads. So imagine that
premium advertisers contractually obligate their publishers to set
f-capping at 0/24 for DNT:1 users (this means that the premium ad would
never be shown to the DNT:1 user). To play the end game, if DNT:1 signals
were ubiquitous on the Web, the overall value of "free access" publishing
would go down and I believe there would be a rapid proliferation of
payment gateways in response (the money to pay for content and innovation
has to come from somewhere). Once again, enter the new digital divide
(where the 'haves' pay for access and the 'have nots" are denied access,
based on financial ability to pay), courtesy of this working group, IF we
don't get it right.

Chris Mejia | Digital Supply Chain Solutions | Ad Technology Group |
Interactive Advertising Bureau - IAB



On 7/11/12 1:15 PM, "Tamir Israel"<tisrael@cippic.ca<mailto:tisrael@cippic.ca>> wrote:

Chris -- I personally found your explanation very useful so thank you.

On 7/11/2012 3:27 PM, Chris Mejia wrote:
Advertisers have plenty of
reasonable business reasons to require f-capping in their contracts:
i.e.
a) not annoy consumers with overdelivery when such annoyance leads to
negative advertiser brand association, and b) not needlessly waste ad
impressions and money on serving ads over and over again to users who
have
opted out of the value exchange in the first place.
It's not clear to me that selecting a DNT-1 means opting out of the
value exchange. The very fact that you need to F-cap those who have
chosen to send a DNT-1 seems to imply that these impressions remain
valuable, at least to some extent (or, I imagine, no ad would be served
at all and we need not worry about annoying users with repeated
exposures or maximizing ROI).

Best,
Tamir
Received on Wednesday, 11 July 2012 23:50:36 UTC

This archive was generated by hypermail 2.3.1 : Friday, 3 November 2017 21:44:52 UTC