Facebook, Attribution and Cookies

I thought I’d put this out to the community since I would love to engage in conversation around this.

At the agency, we have recently seen significant positive performance on a FBX campaign; performance as measured by an online sale (lets keep it vague).

I have been thinking about this and emailing with a few folks about why we might see such stellar performance.

I think I know the answer but want to run it by all of you, to help me think it through.

Facebook is used by over one billion people.  Many of the users of Facebook keep it open in a browser tab all day but it might not be “in view” most of the time.  However, there are consistently six ads in the right rail, all of which consistently update (and theoretically, drop cookies).   Using Ghostery, I see that DoubleClick has a tag on my Facebook newsfeed as I write this.

Is Facebook the new AOL Instant Messenger or Pop Under where it persistently is refreshing cookies all day long and taking credit for conversions?

Triggit recently ran a study where they converted 36% more re-targeted users than Google Display Network, Rubicon, Admeld or Pubmatic.  Is this because they have 36% more reach (I’m not sure if they do, I’m not logged into comScore at the moment).

In a world that is using last view/click attribution, then this could be a real issue for measurement.  If you are using a more advanced attribution method thru VisualIQ, Adometry or Encore Metrics (amongst others), hopefully it get teased out.

Just thinking out loud- lets discuss.  Leave a comment or email me thru the contact form.


  • http://www.triggit.com/ Christina Park

    [Disclaimer, I work for Triggit]. I think this is a valid point and that advertisers should definitely be working with a partner that can offer sophisticated attribution and optimization methods (e.g. A/B testing, frequency capping, etc.) since as you correctly point out, over a billion people are on Facebook. Here’s what’s worked best in our experience when thinking about FBX attribution: 
    1) I agree that viewthrough attribution on Facebook does have the potential for abuse if you’re an advertiser running 100% last view through attribution, but I don’t think any advertiser uses that kind of attribution model.. and if they do, I wouldn’t advise it!
    2) In a control test (which was used in the study you referenced), users who see FBX ads convert at a higher rate than those who do not. This was even true during a click-based test, which is nice since you don’t’ really have to worry about a “constant refresh” problem. Not only that, but doing last-click attribution analysis showed that ROAS is higher than other exchanges, the CPAs are at a fraction of the cost, and more users are likely to convert after they see and click through from an FBX ad.
    3) Again, can’t stress enough – advertisers should work with a transparent, data driven DSP that can at least put in frequency caps on an hourly or daily level and carefully measure their results to ensure they are getting their money’s worth.

    Don’t hesitate to email sales@triggit.com if you want to follow up with more questions!

  • http://www.facebook.com/msgmsg Matthew Goldstein

    Excellent point.  Guess this is why we need fractional attribution. 

  • http://www.linkedin.com/in/anthonyyi Tony Yi

    the cookie issue likely exists but not sure to what extent it affects performance.  depends how the campaigns were set up by triggit but the optimizations described below seem very standard/basic. 

    better performance in general not surprising considering facebook (single premium publisher) is being compared to ad exchanges/ssp’s with thousands of publishers – many longer tail.   also factoring in the fact that fbx inventory is cost efficient is a huge signal. 

    not a big fan of A/B tests but huge, huge believer in attribution methods and understanding the value of top/mid funnel tactics.  something very often ignored or diluted into retargeting metrics/attribution today.  big problem.  keeping a lot of great social, branding and interest spend under value.

  • Chris

    Facebook ads often do not mention the Brands name in the ad.  In this case there is clearly nothing to be attributed to “Brand Lift” or “Post View Effect”.  

  • http://twitter.com/johnbattelle John Battelle

    Hope you get an answer Darren, if you do, let us know!

  • http://twitter.com/CountXero Jeff Ferguson

    It’s definitely not a crazy idea, but comparing it to the pop under
    that refreshed cookies is a bit unfair. Facebook is legit, we and
    everyone we know use the hell out of it and if you advertise there,
    it’s probably at least a part of the attribution equation; the pop
    under thing was pure scam and not worthy of comparison.

    Attribution is a funny discussion; most claim it’s a broken piece of
    Internet marketing tracking, however, it’s something that is easily
    found in e en the freest of analytical software. Where people truly
    seem to be throw is on the question of “how much” credit each channel
    in the process deserves. Frankly, I’ve always said, if it shows up, it
    gets credit and keeps getting funded, no matter if its the first or
    the last piece of the puzzle.

  • http://twitter.com/HowieGoldfarb HowieG

    Over the summer comScore said that the average person in the US spends only 13 minutes using social media a day. I have observed the 15/80 rule. 15% of the users are heavy users creating 80% of the content. The rest are passive or just have accounts. I would believe Facebook getting credit of it is left open and I then went to buy something elsewhere and it gets credit. The management is a bunch of slimeballs under pressure to get their stock price up.

    In my view the only credit facebook should get is when I click on the ad it takes me somewhere and I buy something. 

    Best solution is using Firefox and blocking all the facebook ads like I do. 8)

  • http://twitter.com/mishasilin Mikhail Silin

    Any 3rd party ad network/exchange, even Facebook, will claim 100% credit for every conversion they see as they don’t have the complete picture. The best way to look at conversions is from a multi-attribution perspective, using an unbiased 3rd party platform such as Convertro. Using Convertro’s attribution modeling to split credit amongst all the views and clicks involved in each conversion, it becomes much easier to calculate the actual Cost per Acquisition, which is what we at Convertro did with Gamefly a while back. http://www.internetretailer.com/2012/12/17/gamefly-finds-customers-through-facebook-ads

  • http://www.facebook.com/profile.php?id=803701 Seni Thomas

    Is there an apples-to-apples report of click-conversions vs. view throughs across fb/google/etc?  If it is really heavily weighted on view-throughs then you have your answer, or at least discount it by a fair percentage based on the ratio of clicks to views on average from the other players. In general aren’t view through discounted anyway on the agency side?  With all the below the fold, drop the last cookie on page load tactics that most everyone still employs.

  • Paul Pellman

    The challenge for marketers re: accurate reporting will only continue to get harder/worse as 1) volume of digital media that marketers buy increases (meaning that every targeted user sees an ever increasing stream of media from the advertiser) and 2) additional “walled garden” media channels crop up (in the last few years — Facebook and Twitter; soon to be followed by Pinterest, etc.)… 
    But, it’s hard to blame the publisher — frankly they want to be able to prove that their media is driving value.  And, unfortunately they can only use the data they have — which is typically just what they sell to the advertiser and any pixel coverage/conversion data they can get.  Typically, the advertiser just isn’t (rightly so) going to give the publisher deep/granular data on the other digital media they’re buying. 

    That’s where advanced attribution can help: It helps an advertiser use and cleanse all of the data to figure out what parts of their media campaigns are ACTUALLY driving conversions. This is where last click attribution methods fall short.

    We’re starting to see savvy publishers embrace their advertiser clients that use advanced attribution — basically, it allows the publisher to 1) know exactly where they stand and 2) use insights that the advertiser provides to optimize/provide more of the right/performing media to the advertiser.   That’s when we create a real win-win…

  • http://twitter.com/markmannino Mark Mannino

    FBX is the new AIM game – I’ve heard media sales guys tell me this.  The old AIM game was when people ran little 120×120 ad units in the AOL Instant Messenger application which cycled ads, dropped cookies but was mostly in the background.  I ran campaigns on AIM and never ever saw a click but saw lots and lots of view-through credit – tons of it.  I also saw campaigns I ran lose a large % of their last impression credit the moment a client turned on an AIM campaign.  Luckily for all of us, AIM is largely gone.

    But now Networks and DSP’s without access to FBX are losing the retargeting budgets of their best clients to those with access because those with access can drop tons of retargeting cookies all day long and win the last impression.  Christina from Triggit says she doesn’t know any advertiser still using last impression attribution – um, what?  I still see most using it. Sure lots are using advanced multi-touch attribution but most advertisers just use DFA which is last-touch.  

    It’s not all baloney.  And it’s not just working in retargeting. Lots of advertisers and agencies are doing interesting studies and analysis showing FBX does have benefits.  But the sheer scale also drives this attribution game. People sample facebook all day and if you’re touching consumers consistently throughout the day at high frequency the odds are in your favor to be the last impression before a conversion event.

    And it’s not just the scummy guys who are doing it – to some extent, if you’re using a basic ad server you can’t help it.   If you’re not using an advanced attribution system, the best solution might actually be to keep it out of your ad server altogether and evaluate it separately – then you can see what all your other channels are doing and also try to figure out if FBX is doing anything.

    I tweeted about this back in November actually – https://twitter.com/markmannino/status/274138076748607490