How Advertising Drives Business Results: An Attribution Primer

Mar 15, 2018

Despite the advancements in machine learning and predictive analytics, many marketers struggle to understand the value that each element of their marketing plan brings. In a 2017 AMA study of marketers about marketing ROI, only 25% felt their team had the right tools and processes in place, down 8 points from the previous year.

To answer this question, marketers utilize consumer research, category insights and various forms of marketing analytics, including attribution modeling.

Attribution Modeling

Multichannel attribution modeling is defined as “the process of assigning credit to different customer touchpoints (ad views, clicks, video views, etc.) along their path to conversion.”¹  This assigned credit becomes a way to set performance metrics along the customer path to purchase. Attribution modeling can help you optimize media placement, understand the customer journey at a deeper level, and justify marketing spending.

The Customer Journey

Fundamental to attribution modeling is a strong understanding of the customer journey. For example, let’s consider a hypothetical path to an online sale at an ecommerce website. The path might look like this:

Step #1: See ad on YouTube.

Step #2: Seach on Bing.

Step #3: See Facebook page.

Step #4: Go to ecommerce site.

Step #5: Make purchase.

In a perfect world, the attribution model would identify the value of each step in this consumer’s journey, based on the value of the ultimate sale. When enough consumers had completed their individual paths to purchase, the credit would be added together for each medium, providing a numerical benefit that could be compared to the cost.

One notable characteristic of this path is that it all occurs online. However, for many businesses, multiple consumer touchpoints online can drive to a conversion offline. This leads us to a trickier attribution challenge.

Offline vs. Online Conversions

Let’s consider, as an example, another hypothetical consumer journey.

Step #1: See product photo when browing Pinterest.

Step #2: Hear radio ad while running errands.

Step #3: Drive by outdoor billboard.

Step #4: See store and drive in to make a purchase.

This example path not only has an offline conversion, but includes far-less-trackable offline media. Again, in a perfect world, each medium used would be assigned a value proportionate to its role in delivering a sale. This is very difficult, not the least because we don’t know when a consumer sees/hears a particular offline advertisement.  

Top-Down vs. Bottom-Up Attribution

In an effort to understand the value of consumer touchpoints, many marketers utilize Marketing Mix Modeling. This is based on analyzing all the inputs (e.g., media, promotions, competition, weather, economic factors) and developing statistical models to understand how each variable impacts the output, which is usually sales. This approach originated before the dawn of the Internet as a mathematical methodology to determining marketing value. Sometimes it is now referred to as “top-down” attribution, because it starts with all the weekly marketing inputs and associates them with sales. By contrast, the advent of digital media has created opportunities for “bottom-up” attribution, which begins with sales and traces their origin back through multiple consumer touchpoints.

Given enough time, expertise, and money for investing in measurement, the best option is a combination of top-down and bottom-up approaches. Top-down is best for strategic media allocation choices, and the best marketing mix models include software designed to predict the most profitable option from various user-generated scenarios. Top-down models are also powerful for understanding outcomes other than sales, such as brand affinity, and inputs such as salesperson staffing. Bottom-up models allow for more frequent optimization since the data they use are far more granular and timely.

Challenges in Attribution Modeling

It should be clear by now that there is no perfect approach to determining the impact of each individual medium on store traffic, brand health or sales. Here is a brief list of many of the concerns that should be considered in implementing an attribution model:

  • Online vs. offline conversions
  • Need to optimize for fast-moving businesses
  • Complex consumer journeys
  • Offline media
  • Missing/poor data


In summary, calculating the effectiveness of a multiple-channel marketing plan driving offline sales is complex and challenging. There is no perfect, single solution. But utilization of any of the models referenced here is an improvement over gut feel.  

Contact us today to learn more about attribution modeling and how to use it to your brand’s advantage.


—Julie Pahutski, Curiosity Advertising