A Sound Check For Digital Attribution

By
  • Facebook
  • Google Plus
  • Twitter
  • LinkedIn

marcrossen“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 Marc Rossen, senior director of digital attribution at MarketShare.

I am an audiophile. I enjoy music, especially the quality of music.

As an audiophile, I have come to appreciate the importance of nuance in an ecosystem that can best replicate reality. As I write this, I’m listening to a Studio Master recording of Miles Davis’ “Kind of Blue.” It was transferred from the original analog tapes to a 24bit/192Khz digital audio AIFF format. It’s being played through my solid-state MacBookPro via USB to an external digital-to-analog converter (DAC), which is sent to a headphone amp and then to my Sennheiser open-back cans.

You probably think I’m nuts. I may be a little nerdy, but if you were in my office right now, you’d feel like you were at Columbia Studios, circa 1959. It is reality replicated as best as I can afford.

What does this have to do with digital attribution? Simply this: All those technical jargon acronyms above, such as 24/192 and DAC, have important places in a value chain. If one is weak or not matched correctly – replacing a 24-bit digital file with a MP3, for example – everything breaks down. It once sounded real, but now your gut is telling you that something is missing.

Digital attribution has plenty of technical jargon, too, such as predictive models, sample errors and selection bias, to name a few. These technical terms are meant to help describe a consumer reality. How, for instance, did multiple touch points in a sequence of events influence a consumer to buy a product or perform some high-value brand interaction? The challenge for our industry is that we are too wrapped up listening to MP3s and have forgotten what music should really sound like. The value chain is being crippled by weak links.

Consider two examples: offline factors, like TV, and view-through display banner impressions.

In the case of offline TV, say you are running a display campaign with a significant increase in spending as part of a larger corporate media campaign. Results come in weekly with CPAs below your goal despite a larger reach. You have no idea how much that corporate TV buy is causing cross-channel influence on digital and those display banner ads. Are the display ads driving ROI? Probably, but at the ROI you are seeing in your campaign data, definitely not. The way to solve this is by marrying top-down mix modeling techniques with consumer-level analytics. By fusing these, we can illuminate the effect of offline channels like TV at the consumer level creating a much truer picture of digital’s value in the mix.

As we know with view-through display, most display banner ads are not clicked on and even if they are, the click metric is not very usable. Often consumers see these impressions and, at some point after multiple cross-channel touches, end up on the brand’s website several times before purchasing. What happens to all those impressions that drove a website visit but not the final sale? Do you throw them out or account for them, and if so, how? The solution is to use your brand website data and estimate engagement scores with high value content and use these as inputs into a model.

These are two examples in the sea of attribution challenges we all face in our quest to better represent the reality of the consumer journey to purchase. Missing pieces like the offline impact of TV or view through impressions can severely weaken the attribution value chain and give you a false picture of reality.

Here are my key takeaways:

1. Ask your partners simple questions: For example, how do they handle attributing display remarketing impressions vs. email? Does it make sense to you?

2. External factors: Does the approach account for nondigital factors, such as TV or even economic data?

3. Do your research: Look up basic modeling methodologies on Wikipedia and educate yourself on key industry terms. A little knowledge can go a long way.

4. Gut check: Is the pitch from your partners so complicated or glossed over that your gut tells you something missing?

There are multiple ways to approach digital attribution and many great partners in our ecosystem. But it’s on you, the client, to look deeper and think about how well reality is being replicated.

Ensure your digital attribution value chain extracts as much fidelity as possible. It’s complicated and can be a frustrating process but the time spent can save or make your company millions of dollars, leaving you looking like a digital audiophile.

Now, back to some more music.

Follow Marc Rossen (@marcprossen), MarketShare (@MarketShareCo) and AdExchanger (@adexchanger) on Twitter.

  • Facebook
  • Google Plus
  • Twitter
  • LinkedIn

Email This Post Email This Post

By on at

One Response to “A Sound Check For Digital Attribution”


  1. Huayin Wang says:

    Good article, Marc! I particularly like the questions you asked; I also feel it'd be great if you can add one more:
    Ask your partners if they are hiding the attribution methodology behind the claim of proprietary IP or having them in the open for public validation.

    I am completely distrust anyone who make claim on proprietary methodology about attribution.

    Now you can turn on your music.

Leave a Reply