Attributing Behavior 'Wholisitically'

by , Dec 4, 2009, 3:45 PM
  • Comment (5)
  • Recommend (15)
Subscribe to Data and Targeting Insider

At this summer's OMMA Behavioral and this fall's OMMA AdNets conferences, references to "attribution" popped up frequently, as marketers bemoaned their inability to successfully track where their buys were having the most effect in a campaign. Unless you know whether that direct mailing piece or that out-of-home exposure, that portal banner ad or that email blast was responsible for sending people to your Web site or to the store, how will you know how to manage your marketing mix going forward? Success, like science, needs to have repeatable results.

I spoke recently with one company trying to solve this problem by stitching together a holistic view of the customer's online and offline media exposures and their subsequent conversions. According to Visual IQ president Anto Chittilappilly, the company pulls all potential data from TV viewing, radio, out-of-home, e-mail, mobile, etc. from its software-as-a-service product -- and into one bucket. "We will collect the data from each individual users and show that they have seen this many impressions."

That seems like a tall order, with a lot of potential holes in any map of a user's data profile. Visual IQ has a unique ID it embeds in a pixel or javascript that lets the company collect information about an online ad -- both where it is being displayed and the actions taken by the user. On the offline side, Visual IQ uses a range of technologies to harvest data about ad exposures. The company can use vanity phone numbers and discrete URLs that direct viewers to track specific ad types, which can account for a small percentage of views. Company strategists are starting to work with set top box manufacturers so that the data that comes off more addressable upcoming projects like Canoe will track back to online behaviors.

But most of all, Visual IQ uses online data mining to find correlations between offline exposures and online behaviors. "An ad may be placed at 5 pm, and then you will see a spike on Google around the same keywords as the ads," says Chittilappilly. "We have methods of measuring that correctly through data mining and control groups. When we don't have the data we use sampling and surveys to find correlations. Even in the case of digital you may not be able to get the cookie data so we have to go to data mining to find the correlations." 

Clearly the chain of data across media is still in a work in progress, and part of the model also entails pulling in a marketer's existing customer database in order to know better whom to track in the first place. But Visual IQ believes there are enough pieces of hard data already there to make calculated good guesses that fill in the gaps between an ad viewed and a conversion. When you don't have hard data, you fall back on correlations.

The end result should be marketing and media spend insights that are more informed and actionable than is the norm. "If you can know that people who have seen this particular creative on display are using this keyword on the search side, then you can determine the best mix of creative," says Chittilappilly.

Ordinarily it is the last click that gets credit for acquiring a customer, which is why search continues to get such a lion's share of media spend. But only a holistic view of the entire chain of communications about a product with a customer can give an assessment of what messages were triggering what behaviors. "We can tell all of the engagements that a user has had in the past and give appropriate credit to those exposures," he says.

Chittilappilly claims that with a more complete understanding of a user's history of media exposure and interaction, the marketing spend can not only be apportioned more effectively but the media mix itself can be optimized for better effectiveness. "What time do you put which creative on display? That has to be in harmony with the emails and the keywords. It is every channel working in synergy. We promise about 20% to 25% lift in ROI on media spend," he says.

Classic behavioral tracking technologies simply use a publisher's data to find patterns in a user's behavior that can be exploited by marketing. In some sense the holistic approach Chittilappilly and Visual IQ create reverses that polarity and uses the large picture of a user's cross-platform behavior to demonstrate the value of less measurable branding campaigns on conversions that usually get attributed to search and direct marketing.

Holistically, the display ad publisher is playing a much bigger role than previously thought in initiating the chain of actions that leads to conversion. "We have brought back a lot of publishers who have been sidelined," he says. "You can see which publishers are causing lifts in search keywords."

Search won the first round of digital ad spending not just because it worked and was measurable, but because it was also among the first things to be well measured online. When that last click is accurately placed within a larger context in the history of media exposures, then the basic math of marketing mixes could shift.    

0 comments on "Attributing Behavior 'Wholisitically'".

  1. Chris Nicholls from asiesta4u
    commented on: December 5, 2009 at 12:30 p.m.

    I did some work with a company called Integration-imc who also did a lot of work for P&G & Nissan.
    They got as close as I have ever seen to finding the holy grail of marketing.
    In my view it was about 95% accurate depending on how smart the strategic planners were who ran the programme.
    It literally helped you compare the impact of any form of paid for or non paid for communication on brand preference and purchase !
    Not a cheap exercise but worth checking out.

  2. Howie Goldfarb from Blue Star Strategic Marketing
    commented on: December 5, 2009 at 3:54 p.m.

    Another great enlightening article Mr. Smith! Sadly for Visual IQ I don't allow scripts to run on my FireFox browser because of Windows susceptibility to viruses and malware.
    And I only allow Ads on very specific websites that I feel should get money providing me with free content (such as mediapost). And per FireFox there is over 700,000 downloads of ad blocker plus and over 300,000 for No Scripts per week! And while I might feel that websites that provide free content should be allowed Ad Revenue because of the invasions of privacy levels that sometimes go beyond what I feel is ethical (nevermind the virus and zombie control malware), I think having a countering check-balance is a good thing.

  3. Aron Anthony from Digitas
    commented on: December 7, 2009 at 10:51 p.m.

    Excellent article Steve. Most attribution technology providers that I have come across just divide the credits equally and attribute to all clicks. Nice to hear about somebody does this more accurately. Would like to know how Visual IQ deals with cookie death. There are ways to get around the issue Howie pointed out. Anybody have a list or comparison of of all the attribution technology providers?

  4. David Williams from Neo Ogilvy
    commented on: December 7, 2009 at 11:20 p.m.

    Aron, I watched Anto Chittilappilly speaking about cookie deletion at ad:tech-nyc. Apparently they have some kind of compensation built into the modeling to tackle the cookie deletion issue.
    Forrester Research has published a wave report just recently. It compares the interactive attribution technology providers with all details. Normally it costs money to get such research reports. But here is link I found on Visual IQ’s website. http://www.visualiq.com/forresterwaveoffer.php
    Hope that help

  5. Michelle Cubas from Positive Potentials LLC
    commented on: December 11, 2009 at 11:17 p.m.

    Steve, this is a brilliant comment and nails the entire issue about marketing.

    "Holistically, the display ad publisher is playing a much bigger role than previously thought in initiating the chain of actions that leads to conversion."

    Marketing Voodoo™ (I call it) cannot predict when, where, what, a person will look out. Plus, are they a tactile person, visual, scratch-n-sniff type? Then, the trick is to understand how they process the information.

    We could do a weekend retreat on this topic. It would be a terrific collaboration as I do so much work around styles and communication. Thanks for the brilliance.

Leave a Comment

Sign in to leave a comment. Don't have an account? Join Now

Recent Data and Targeting Insider Articles

» Data and Targeting Insider Archives