De-mystifying Marketing Attribution: why is it so complicated and how to make sense of it?

During my user research on marketing analytics, one of the most controversial topics that arose was attribution. When I broached the subject, the responses ranged from “OMG, how long do you have to listen to all my gripes” to “I gave up on it a long time ago.” In this article, we will explore various marketing attribution models, their strengths, limitations, and how they can be applied to optimize your marketing efforts. By the end of the article, the reader should have a basic understanding of what it is and when to apply specific models.

According to Wikipedia, a marketing attribution model is

“the identification of a set of user actions (“events” or “touchpoints”) that contribute to a desired outcome, and then the assignment of a value to each of these events. Marketing attribution provides a level of understanding of what combination of events in what particular order influence individuals to engage in a desired behavior, typically referred to as a conversion.”

Well, that sounds easy enough. Or is it? Let’s follow the journey of “Bored Bob” to see why it’s complicated.

• On his way to the office, Bored Bob sees an ad for a new gaming console on a giant billboard.
• While browsing a gaming website on his Android phone, Bob encounters a display ad for the gaming console. Bob’s phone only has his personal email.
• Intrigued by the ad, Bob clicks on it, leading him to a landing page where he explores the console’s features and specifications. A cookie is placed on Bob’s phone to track his interactions throughout the customer journey.
• Bob signs up for an email newsletter to receive updates on the console’s release and exclusive offers using his secondary email address.
• When browsing on his work laptop, Bob utilizes incognito mode while browsing, making it difficult to track his activities accurately.
• Bob then forgets about the console for several months, until…
• While watching gaming-related videos on YouTube, Bob encounters a video clip showcasing the console’s gameplay and features.
• Retargeting ads follow Bob as he visits various social media platforms, reminding him of the console and enticing him to make a purchase. Finally, after receiving a promotional email campaign with a limited-time discount offer, Bob decides to purchase the gaming console.

Bored Bob, generated by Bing Image Creator

Between the display ad, search, YouTube, social media, and email, which one should get the “credit” for enticing Bob to buy the console? The display ad (the first digital encounter)? The email campaign (the last encounter)? The YouTube video (the most time spent on a single action)? Social media (the most time spent on combined actions)? All of the above? Chances are, if you asked Bob, he wouldn’t have a clue either! The challenge Bob faces can be summarized in these points:

A. Complexity of the customer journey and multiple touchpoints:

• The customer journey is no longer a linear path. Customers interact with multiple touchpoints across various channels before making a purchase decision. Attribution models must account for this complexity to provide accurate insights.

B. Limited visibility into offline and cross-device interactions:

• With the proliferation of offline and online touchpoints, tracking and attributing conversions across channels and devices can be challenging. Marketers need to bridge the gap between online and offline interactions to gain a holistic view of customer journeys.

C. Inaccurate or incomplete data collection:

• Data quality and accuracy are critical for effective attribution. Incomplete or inaccurate data can lead to misleading insights and flawed attribution models. Marketers must ensure comprehensive data collection and proper integration to overcome this challenge.

D. Privacy concerns and regulatory constraints:

• Increased awareness and regulations around data privacy pose challenges to tracking and collecting customer data. Marketers need to strike a balance between respecting privacy rights and obtaining data necessary for attribution.

Now, let’s look at the most used attribution models and see what they do.

I. Single-Touch Attribution Models

A. Last-click attribution:

• The last-click attribution model attributes all credit to the final touchpoint that leads to a conversion because it is the only one that we know for certain has a role in the conversion. For Bob, the last touch attribution would be the direct email.

B. First-click attribution:

• “First impressions last,” and that’s exactly what the first-click attribution model does. It assigns all credit for a conversion to the first touchpoint a customer interacts with. It is suitable for understanding the initial touchpoint that captures a customer’s attention and initiates the conversion process. In Bob’s case, he is attributed to the display ad because it is his first digital encounter with the console.

II. Multi-Touch Attribution Models

But how can we only count one point of contact when Bob has clearly interacted with so many places? It wouldn’t be fair to all the other channels! Let’s call the fairness police and rectify this issue. Enter Multi-Touch Attribution.

A. Linear attribution:

• The linear attribution model distributes equal credit to each touchpoint in the customer journey. It acknowledges the contribution of all touchpoints throughout the path to conversion. For instance, given that Bob interacts with a display ad, then clicks on a search ad, and finally converts through an email campaign, each touchpoint will receive an equal share of the credit.

B. Time decay attribution:

• Should all touchpoints be considered equal? What if the display ad was a REALLY long time ago, and Bob doesn’t remember it at all? The time decay attribution model assumes that people forget. The longer the time, the less people remember. Therefore, the time decay model attributes higher credit to touchpoints closer to the conversion event while gradually reducing the credit for earlier touchpoints. In other words, touchpoints closer to the conversion have a more significant impact. For example, Bob engages with a display ad, then clicks on a social media ad, and ultimately converts through an email campaign; the time decay attribution model would assign higher credit to the email campaign.

C. U-shaped (position-based) attribution:

• The U-shaped attribution model allocates higher credit to the first and last touchpoints, recognizing their significance in the customer journey. It attributes the remaining credit evenly among the touchpoints in between. For instance, Bob encounters a display ad, interacts with paid search ads, watches YouTube videos, and finally converts through an email campaign; the U-shaped attribution model would assign significant credit to the display ad and the email campaign while distributing the remaining credit evenly between every other interaction that happens in the middle.

Okay, that all makes sense. I’m still following the multi-touch attribution models. They all acknowledge that all interaction points have a contribution to Bob’s eventual conversion. However, they differ in their beliefs about which one is the most important. Some say “First impressions last,” some say “(last) action speaks volumes,” and there are combinations in between. They all somehow make sense, so how do we know which one is correct? Is there a way to combine them? Is there a cleverer method to assign credits? Enter Algorithmic Attribution Models.

III. Algorithmic Attribution Models

These models follow the original intuitions but also allow much greater flexibility in terms of assigning weights based on how ‘influential’ a particular channel or activity is. They incorporate advanced statistical methods and machine learning algorithms to come up with the most ‘right’ answer. Several of these models exist, and some are vendors’ proprietary models. They tend to be more sophisticated but are also harder to understand and therefore are often seen as ‘black box’ solutions.

Choosing the Right Attribution Model

Well, that was a whirlwind of information. After all of that, which attribution model should you use? The answer is, of course, “it depends.” Selecting the appropriate attribution model depends on various factors, including the nature of your business, marketing objectives, and customer journey. Consider the following when choosing an attribution model:

• Complexity of the customer journey
• Channel mix and touchpoint variety
• Business goals and KPIs
• Available data and measurement capabilities

Matching the model to your business goals and customer journey:

Analyze your customer journey, touchpoint interactions, and desired outcome. Choose an attribution model that aligns with your specific business goals and accurately reflects your customers’ decision-making process.

Importance of testing and experimentation:

Attribution models should be continuously tested and refined to ensure their effectiveness. Conduct experiments and compare results across different models to gain insights and optimize your attribution strategy.

In conclusion, marketing attribution is a complex and challenging aspect of modern marketing. We have explored various attribution models, from single-touch to multi-touch and algorithmic models. Choosing the right model requires considering factors such as the customer journey, marketing channels, business goals, and available data. By continuously refining and optimizing your attribution strategy, you can make data-driven decisions and maximize marketing effectiveness. Embrace the challenges, experiment with different models, and leverage data insights to drive success in your campaigns.