Changing Lanes: Solving the Decade-Old Problem of Cross-Channel Ad Attribution

By
  • Facebook
  • Google Plus
  • Twitter
  • LinkedIn

kim-reed-perell“Data Driven Thinking” is written by members of the media community and contains fresh ideas on the digital revolution in media.

Today’s column is written by Kim Reed Perell, CEO of Adconion Direct.

Recent debate surrounding deficiencies in today’s media attribution models is a positive indication that the industry is finally ready to move forward from its antiquated, decades-old measurement system. In theory and in concept, most agree that our rapidly changing media landscape demands a more dynamic, comprehensive value system. Yet we are still too far from a “perfect world” to realistically apply fractional attribution models across all types of media.

Ad attribution is not a new concept; for years, industry leaders have analyzed exactly what it means and how the concept can reach its potential. But now that the industry is ready to take this discussion to the next level, I’d like to address the real roadblocks preventing us from switching to a new system, discuss why these inefficiencies still exist in 2013, and propose a realistic solution to start improving them.

For most companies and digital marketers today, fractional attribution across multiple touch points and channels – or the idea of distinguishing the impact that each touch point plays in the success of an advertising campaign – is simply an academic dream. In reality, advertisers are still struggling to even recognize view-based attribution credit, which six years ago became the next advancement above last-click attribution.

This lack of progress is not advertisers’ fault; on the contrary, the reluctance to adopt a new attribution model easy to understand.

Common sense tells us that any media touch point has influence, good or bad, on a potential consumer and their purchase decision. However, in the absence of clear impact – a way of determining exactly how much influence each touch point has toward a purchase – it’s easier for brands to ignore that influence, in terms of attribution, and stick with the status quo of a last-view or last-click model.

Advertisers still have every right to be skeptical.  If the industry can’t convince them to attribute value toward a single banner in a single channel, from consumers viewing or clicking that one particular ad, convincing them to buy into fractional attribution across all touch points will be very challenging.

But this is where we sit today. We know intuitively as marketers that multiple media touch points should all be connected and attributed to the performance of an ad. But nobody has put together a compelling enough case to convince advertisers or their agencies, at least in any significant scale, of the right model. The industry has yet to make fractional attribution the default; last-click attribution continues to hold that title.

There are two primary hurdles to convincing executives, decision makers and brands of the need, value, and potential of fractional attribution: 1) the need to centralize disparate distribution channels and media buying onto a unified platform; and 2) a way to effectively track and report the impact all touch points have on a campaign using common industry metrics.

Recently, Harvard Business Review published a thought-provoking article titled “Advertising 2.0.” In it, the authors describe current digital marketing efforts as occurring in “swim lanes,” with media across different platforms running parallel to each other but never crossing over into neighboring lanes. Yet in today’s market, this “swim lanes” metaphor exists not because media buyers are working right next to each other without communicating or strategizing – or even because the technology isn’t there to merge them together scientifically – but because the efforts aren’t even happening in the same pool.

Many advertisers still work with specialists or platform-specific providers, a hangover from 2010 and LUMAscapes of old. The result: the piecemeal aggregation of distributed media, execution partners, ad technology systems, agencies, and media channels – on which most advertisers rely for a comprehensive media plan – restricts even the most forward-thinking agencies. How are we to expect media planners to swim in the same pool, much less in parallel lanes, when media is still segregated and disjointed with respect to display, mobile, social media, video, and email?

Ruling out and ignoring the immeasurable touch points (beyond the digital landscape, what about the impact of offline branding like billboards?) is incomplete at best. I think most of us know it. The challenge is getting us collectively past this way of thinking, because today we’re far from where we can be as industry. Different media-buying partners across different platforms, using different reporting and tracking systems, equate to a challenging if not impossible cross-channel dream.

However, unifying digital spending is a way for most advertisers to begin aligning their attribution capabilities. Simply put, if your media campaigns aren’t served, managed and distributed from a single platform for a single user, you can’t know the impact of one touch point to another for any given user and therefore can’t make meaningful attribution models.

What are some other realities and solutions? More importantly, what are meaningful solutions that advertisers can use today?

First, companies should be weaning off the last-click model within a single channel, a channel where spend and ad serving is unified to begin with. Next, companies should take small steps toward full-funnel attribution even within that single channel. Start with a smaller marketing effort or campaign that doesn’t have to disrupt the current business model. What were all the possible touch points that simply led a new user to land on your site in the first place? Look at these new users and how they use direct navigation, in addition to what they click, and start to create attributes around them. This is one example of how to gain transparency into user-action impact to prove that viewing an ad has an effect. This will lay the groundwork for larger view-based attribution, which will then pave the way for fractional- and then cross-channel-based credit.

Is my suggestion a perfect solution for advertisers still buying media and analyzing impressions the old way? No, but it’s a viable start. Are there other variables that aren’t being accounted for? Yes, of course. But this approach would lay the groundwork for demonstrating that a last click isn’t the only media result that drives action, and it would begin to wean us all off our decade-long crutch. These small changes will start proving that some media creates and drives real impact, while some doesn’t. And, as they do, these changes will provide the initial shove to move us out of our collective pools and into a broader media ocean, where we can effectively confront the tides pushing us away from true media efficiency.

Follow Adconion Direct (@Adconion_Direct) and AdExchanger (@adexchanger) on Twitter.

 

  • Facebook
  • Google Plus
  • Twitter
  • LinkedIn

Email This Post Email This Post

By on at

5 Responses to “Changing Lanes: Solving the Decade-Old Problem of Cross-Channel Ad Attribution”


  1. Scott Krauss says:

    Hey Kim, fantastic read. I'm actually in the middle of writing a blog post that speaks to your underlying point here. Which is....marketers in seek of attribution will need to leverage a "unified platform"....the "end to end" tech stack so to speak, It's been an interesting past few weeks as the value prop for this solution continues to gain momentum and it seems that the large tech/data/crm players are well aware of this and rapidly trying to build this solution. In my humble opinion we're about 6-8 months away from some of the major players having this ready to go to market.

  2. Kim, you make some very interesting points. I personally think that the industry is already here but marketers need to be more educated on the tools available to them. The confusion created by solutions that only attribute conversions to a first or last touchpoint can lead to misunderstandings about what true attribution actually does and delivers. Let me clear this up, attribution is about giving credit where credit is due - to each marketing touchpoint that a consumer interacts with from first exposure to conversion. Customer path attribution is able to pinpoint a customer’s behavior step-by-step from the first commercial she sees on the TV, to the ipad she researches the product on, to the laptop she makes the purchase on. By following each step a customer takes between initial ad exposure and conversion, the credit for the final sale can be properly distributed to each touchpoint that impacted the customer along the way.

    Jeff Zwelling
    CEO of Convertro

  3. jeremy says:

    fractional attribution is simply moving the deck chairs around on a sinking ship. it's not even a marginal improvement over last-view. it's basically the same thing.

    all of the models currently in discussion are still based on correlating ad delivery (through data targeting and optimization) to users whom algorithms have deemed likely to convert. this is only part of the equation. what's missing is being able to measure whether or not the impression actually influenced the user to take an action. otherwise, you are just sticking an ad in front of someone who is likely to convert anyway.

    there is only one way to measure whether ads are making a difference. It is necessary to conduct an experiment using a control and an exposed group of users. any other form of measurement is 100% worthless.

  4. John Were says:

    Great points Kim. Jeremy, when you talk about " an experiment using a control and an exposed group of users" would you be able to elaborate? With our advertisers we try to persuade them to run a control ad from one of our other advertisers. This can demonstrate the difference in behaviour between users who have seen the different ads using familar cookie-based measurement techniques; causation rather than correlation. From there we try to determine the value of the ads by mapping against a perceived impact on key metrics such as overall site traffic, transactions, revenue etc. and we then set cookie-based performance goals accordingly. By assuming a % of users don't have cookies on their machines we can calibrate the goals and also buy against these users at lower cost as there is less competition for them. I'd love feedback on this approach.

  5. jeremy says:

    John- This is a good approach. The key is to make sure that the control and exposed groups are structured correctly. Targeting bias is what leads to skewed results. Another issue to consider is that A-B testing is not scalable across an entire media plan with multiple tactics and vendors. The best approach is to employ a causal inference test. This is a method to derive a test and control group without the need for test ads. The original paper on this topic was written in the 1970s I believe and is just now making its way into our industry. I know Mediaplex for example has hired statisticians trained in this type of analytics.

Leave a Reply