I hear it all the time, customers are most interested in understanding their marketing channel attribution. It makes sense, no one wants to stand up in a meeting and deliver the classic marketing disclaimer, “half of our marketing is really effective, I’m just not sure which half”. So, how do we start to get after what activities are truly driving conversion or assisting it?
At the highest level, marketing attribution represents the process of identifying the value of attributes (or put another way, actions) towards a desired outcome (e.g. purchase, lead collection, form completion, etc.)
Adobe Analytics provides the following methods of attributing marketing channels, campaigns, and actions with conversion events:
• First Touch: Assigns 100% of the revenue to the first channel interacted with (can be based on user or visit)
• Last Touch: Assigns 100% of the revenue to the last channel interacted with (can be based on user or visit)
• Linear / Participation: Assigns the revenue in equal weights across all channels interacted with
So what is wrong with these traditional attribution methods?
Inherently, there is nothing wrong with traditional attribution models. Often they are great ways to obtain insight regarding the entry interactions, last interaction, or list of interactions that eventually lead to a conversion. As a matter of fact, I would encourage the use of multiple models depending on the insights you are attempting to extract from an attribution perspective.
The problem is not with the models but the limited amount of models that are traditionally used to obtain insights. In an ideal situation, we would have more sophisticated ways of assigning credit to our interactions.
The good news is there are methods of applying more sophisticated models to your Adobe (or other analytics platform) web analytics data today. Through the use of data warehouse or data feeds from Adobe (Google, etc.), Numeric can help you look at attribution in a completely new way.
The beauty of sophisticated attribution models is that they can be custom built to work anyway you would like them to work. Below are a few common examples of attribution models that extend past the traditional methods.
• Time Decay: Assigns credit based on the distance of time between the interaction and the event
• Position Based: Assigns credit based on the position of the interaction (Position 1 is the first interaction)
• Engagement Based: Assigns credit based on the website engagement created by the interaction
The graphic below shows how the different models effect how a marketer understands the impact of each channel in driving revenue. For example in the First Touch model, Paid Search gets 100% credit for driving the sale. In the Last Touch model, Natural Search gets 100% credit for the transaction. But maybe the Engagement model is a better reflection of how to adjust your spend by channel?
Attribution Model Example
As you work to optimize your marketing spend, the discussion around attribution models will become critical. Most valuable is to have the discussion organizationally and to decide and acknowledge the trade-offs of the model you choose. It can also be useful to consider looking at several different models to see how the insights you are attempting to extract shift with the different perspectives. This will also allow all stakeholders to feel invested in the process and encourage collaboration instead of resentment around which activities should get more investment for delivering the sale.
