Targeting and Attribution – Facebook’s Substantial Lead

6 March 2014

A very very hot topic in digital advertising today is attribution. My definition of attribution: The process by which an advertising campaign measures its influence on consumer behavior. Digital advertising is typically measured by: Ads presented (Impressions), Click Through Rate (CTR), Cost per thousand (CPM), Interaction time (see DoubleClick Data and Top 10 Metrics). Marketers have more data for online advertising than for any other channel, the problem is that people don’t live online. For example, eCommerce sales are around $180B, compared to total Retail sales of $2.4T (excluding Auto, Financial Services and Gas). Similarly Google owns 50% of the digital ad market, with US revenue running at over  $30B/yr, which is just a small slice of the overall US marketing spend of over $500B. The CPG vertical for example is the has the largest marketing spend (P&G $3.2B), but very low digital spend (see Retailer as Publisher).

The marketer’s key “nut to crack” : how does online advertising influence offline behavior? (attributing behavior). Facebook is leading the world in 2 critical areas of advertising: Targeting and Attribution.


Facebook is highly differentiated here, think lasers vs nuclear weapons. Not only can you build a custom audience based upon email, phone, … etc. You can have Facebook expand that to a lookalike audience, or use external data to form a partner audience (consumers that drive a Mercedes, are over 40 and drink OJ). There is no platform on the planet that does a better job targeting. Tech Crunch covered most of this in an April 2013 Article.  Also a consumer privacy group has a very detailed article on issues surrounding facebook/datalogix.


This is where the stakes get much higher, and the facts are VERY closely guarded. Why the secrecy? Perhaps data use is beyond the scope of use agreed to, or at least the “value” of the use has not been realized by the owner of the data. For example the Tech Crunch article outlined how Datalogix used grocery store loyalty card information in custom audience creation (targeting) and attribution. However, Datalogix may not be authorized to use the data in this way (at least for all of the Retail clients).

Lets assume that they have no rights to use Safeway’s data for either targeting or for attribution, how do they get around it? For Targeting: my guess is that they are using a smaller Grocer’s (GroceryX) data to construct an initial data set that Facebook expands (via lookalike). For attribution, they then use loyalty card purchase information to statistically project the performance of the original data set (projecting the purchase behavior of the GroceryX’s loyalty customers on the larger data set).

If this is the case, then GroceryX’s data contributed all of the attribution performance (as well as for targeting). Subsequently the revenue that SV should receive is far above their data’s representation in Datalogix’s grocery macro database. In otherwords, SuperValu (or another unknowing participant) may not be getting paid for the value they are creating.

Regardless of the data use, Facebook is becoming a CPG’s dream channel, far exceeding the performance of anything they have ever worked with (by a factor of 5+!). This is one of the reasons I’m very high on Facebook, and I do own the stock. It may have taken them awhile to figure out targeting and mobile advertising, but they are absolutely killing it today. I believe they could easily grow their CPG advertising 10x in next 18 months.

Purchase Behavior.. Who has it?

There is SIGNIFICANT data leakage going on today. It is a Tsunami that is about to hit every retailer. Data is being used far and above its intended purpose. Another grocery example is what was UPromise, and now SavingStar.  UPromise was an original construct to earn points toward college tuition from SallieMae. Every grocery provided their data to the program so their consumers could participate. SavingStar has tremendous data.. but what can they do with it?  Bank of America’s card linked offer program started to use this data, but the issues of use, ownership and the latency (ex getting credit on day 3) issues persist.

Retailers run a very profitable business in data today. It is core to the current status quo, particularly as it relates to trade spend ($200B/yr). Most retailers are very, very conscious of issues surrounding data leakage. The leading Retail analytics companies (Catalina, dunhumby, Spire, Inmar, ..etc. ) could do wonders in attribution if their data owners would let them.

Purchase Events

Another entity that has purchase data in the US is Argus Information, a Division of Verisk. A little over 10 years ago, Argus evolved as a US bank marketing utility for measuring/targeting cards. Banks send Argus all of their card transaction detail and Argus creates reports for banks (ie Average Customer spend vs competitor in region, average customer balance, …) it was a benchmark service, plus a way for Banks to target Card mailings.  Argus’ former CEO Len Laufler is now running a new data Division at Chase for Jamie.

My friends tell me that Argus has been openly discussing how it can sell its purchase intelligence to non-banks and advertisers (this year). I can tell you one thing for certain, Banks are not cool with this. The head of Retail at a top 3 bank called up Len 2 years ago and told him in no uncertain terms, that the moment they sold their data outside of its intended use they would no longer receive it, and find themselves in front of a judge. The Banks are at risk, Argus is at risk, Consumers are at risk.. if data is used beyond the approved usage. The only way to get this data is with the approval of issuer and consumer.

AdAge had Amex/Mastercard story along these lines in April.   I was also told last month that another source for the data could be Yodlee. As Yodlee’s very first customer (Wachovia 1999) I would say that they have an advantage of customer permissioning. They also have experience in dealing with 3rd party use (Mint, offermatic, …), problem is that it takes time to get the data (customer must register), and there is a latency between transaction, bank record keeping, OFX polling, attribution logic, .

GoogleZave Reciept detail

Quite frankly Google has all of the assets to kill CPG/Retail. Their Zave purchase has put them IN the IBM/Toshiba 4690 OS (run by 16 of 20 top retailers).  Every time I shop at my local Harris Teeter and use electronic Coupons.. it is Google powering a fantastic consumer experience. Customer level SKU information attribution nirvana. They also have a unique content delivery mechanism (targeted incentives) that Facebook can’t match. Manufacturers are not keen to issue coupons to everyone.. they want to target incentives to specific buyers… However Retailers DO want coupons for everyone, unless someone will pay them more to change their behavior. It will take Retailers, Manufactures and Consumer participation to make this all work.. which means tremendous focus (and investment).



5 thoughts on “Targeting and Attribution – Facebook’s Substantial Lead”

  1. Hi Tom,

    Great article above. In principle I do agree with Facebook having cracked the Targeting and Attribution problem. Custom audience and ‘similar to’ audiences are also great, but I am not so sure this works in practice as well as we hope. This I attribute to a lot of incorrect data and personality type attribution. The idea that I ‘like’ a Brand might not really mean in Facebook terms like I support and advocate the Brand. Facebook has a very fuzzy definition of this. For my own business, we use Custom audiences a lot, but I am not yet convinces the data itself is correct. Not the mention the massive contamination for Like Farms. I am sure Facebook is improving this and shall becomes the CMOs dream as you suggest above.

    The BBC did a very interesting report on Facebook likes recently, which highlights my point here:

    Separately, I would love to get your views on some ideas around CLOs. Is there an email I can reach you on?

    1. I think you are spot on here.. as problematic as Facebook is with targeting/attribution.. it is 3-5 years ahead of everyone else.. they are fixing things at an astounding rate. They are onto something very, very big.

Leave a Reply

Your email address will not be published.