This is a continuation from my blog post that began with the basics of campaign attribution and focused on single-touch models. In the below, I elaborate on multi-touch attribution models. To learn more, download our attribution methodology resources.
You’re investigating whether multi-touch campaign attribution is right for you and your company. That is a great sign. You’re tasked with being performance-driven, and you are looking to gain more yield from your marketing campaigns through better understanding of your campaigns’ revenue impact. So, let’s get right to the concepts, the models, the mechanics, and how to incorporate multi-touch into your metrics reporting.
The Allure of Multi-Touch Attribution
As I alluded to above, performance-driven marketers usually begin investigating multi-touch attribution to gain more in-depth insights into the revenue impact of different marketing activities. Marketers look beyond single touch models due to some of the limitations that come with allocating revenue from deals to just one campaign. The limitations of single touch are glaring and are made obvious in the below scenario.
Imagine a marketer’s event generating a lead that eventually turned into a $1M deal for the business. A first touch model ignores all the other engagements and attributes the $1M deal to the event. The timeline looks like the below.
Single Touch Attribution of a $1M Deal
Now let’s see what a pie chart showing revenue per campaign type would look like under this model.
The obvious conclusion: do more events to generate more $1M deals, and if budget is limited, take the marketing money from other campaigns and channels to run more events. Obviously, this could be very detrimental to the business and I’ll get to that in a second.
Now, under multi-touch attribution, the timeline looks significantly different.
Multi-Touch Attribution of a $1M Deal
If we use an even spread model, which allocates the $1M revenue evenly between the four engagements, then each campaign gets $250K of revenue impact credit. Let’s look at the same pie chart but under an even spread model.
This implies that the company should not only emphasize events but to ensure that emails and webinars are a part of the marketing mix in order to optimize for revenue based on what has worked in the past. This can be similarly seen if we compare any single touch model (and not just the first touch model) to multi-touch. The underlying point: single touch models under-emphasize the effect of multiple engagements throughout the buyer’s journey.
The Standard Multi-Touch Attribution Models
Here are the standard multi-touch attribution models:
- Even Spread – allocates revenue amongst all engagements equally
- U-Shaped – allocates revenue to the first and last engagements more-so than any touches in between
- W-Shaped – much like the U-shaped, except this allocates some revenue to middle engagements as well
- Time Decay – allocates revenue to engagements in rank order of how close the engagement was to the deal (essentially an increasing revenue attribution over time along the buyer’s journey)
See below for a fundamental breakdown of each of four multi-touch attribution models, including why they are used and their pros and cons.
Even Spread Attribution Model
As the name implies, even spread is a model that allocates the revenue from a deal across the different engagements in a buyer’s journey prior to the transaction. So, if there are five campaigns engaged, then all five get equal revenue impact attributed to them (so, for a $1M deal, $200K is allocated to each). This is usually the first multi-touch model that is implemented, mainly because it is one of the simplest to implement. Let’s take a revenue pie chart in Salesforce and see what it tells us.
When looking at campaign attribution metrics, the size of the each slice tells us which campaign (type) were engaged with the most in the buyer journeys that led to revenue. Now, the downside to using this model is that this applies equal weighting to all touches, so there will naturally be an over-allocation of revenue credit to unimportant campaigns that happened to be engaging or over-utilized. This is what I call “stuffing the ballot”. For example, if a very engaging email was sent to everyone, multiple times, the email channel might show up on the pie chart as being very important, while a one-time event may have been more important to deals but gets under-allocated revenue credit due to the infrequency of the event. The even spread model could cause the company to mistakenly divert event marketing dollars to email, leading to unoptimized marketing yield.
U-Shaped Attribution Model
This model allocates revenue mainly to the first and last touches of the buyer journey that led to the deal (variants of this model may split a miniscule amount of revenue evenly between the middle touches). Since campaigns that created the first and last touch get the majority of the credit, this model surfaces campaigns that are effective in either generating new leads or influencing last touch conversion (in Salesforce, last touch is defined as the touch prior to Opportunity creation). This model is especially good for companies with a hybrid goal of lead generation and last touch conversion, and even better if their sales are most influenced by activities in the beginning and end of the buyer journey. See the below revenue pie chart and the way one could interpret it under this model.
With that said, the model does not differentiate between first or last. So it’s hard to simply say whether a campaign is good for first touch or last touch. The way to combat this is to then look at a standalone first touch model and a standalone last touch model.
W-Shaped Attribution Model
This model is similar to the U-shaped, except that it allocates more than a miniscule amount of revenue credit to a middle touch. By including a middle touch, this model aligns with companies that want to include campaigns that are effective at some middle point in the buyer journey. Simply put, this model surfaces campaigns that are effective in the beginning, middle, or end of the buyer journey.
With that said, the W-shaped model shares a downside of the U-shaped model in that it mixes together the revenue effectiveness of a campaign along various points of the buyer journey. So a marketer cannot look at a revenue pie chart and conclude that a campaign is effective for a specific point of the buyer journey, only that it’s effective in one of three points (the beginning, middle, and end points).
Time Decay Attribution Model
This model allocates more revenue to campaigns the closer the engagements are to the point of purchase. So, an engagement (say, an eBook download) closer to the transaction date would be given more revenue credit than an engagement that happened long before the transaction.
This model is popular amongst companies that have long sales cycles, especially where campaign engagements become increasingly influential to the buyer journey over time. The downside: this de-prioritizes the first touches, and that is not ideal for companies that need to emphasize lead generation (e.g., growth stage companies that need to build net new names).
Custom Position Attribution Model
Finally, a marketer could utilize a custom position model and weight the touches based on the order in which the engagement occurred along the buyer journey. For example, one of our customers began using a model that valued the first touch twice as much as the subsequent touch. Now, since this is customized, the attribution results from this would be dependent on how you customized the model. In general, engagements at the positions that received higher weightings would be counted the most in the revenue attribution, but that would also depend on the quantity of engagements at those positions. The ability to better match the attribution to the nature of your buyer journey can be very appealing when looking for another model to add to your arsenal. This does start to get complex, and there is a risk of introducing a “self-fulfilling prophecy” bias. The important thing is to continuously validate and utilize multiple models to take a different slice at attribution metrics.
Picking the “Right” Multi-Touch Attribution Model for You
For a comparison table of the above, download this Attribution Model Cheat Sheet, which includes single-touch and multi-touch models. The important factors in deciding on attribution models are: 1) your company and your team’s objectives in trying to assess the revenue impact of marketing campaigns, and 2) the nature of the buyer journey and sales cycle. There is no way to know which model is “right” for your reporting, but there is a right way to go about getting the attribution insights you need. We at Full Circle Insights believe in using the right technology that automates the attribution tracking and allows for enabling of different models at the flip of a switch (plug for our Campaign Attribution product). Secondly, we believe in looking at the attribution results of different models side-by-side and generating a more comprehensive campaign engagement picture.
Outside of methodology, you would want to understand how to implement the models as well as extract insights with second-level analyses. Give us a call and we’re happy to discuss your organization’s unique needs.
Feng Hong is a recovering spreadsheet addict with a background in covering enterprise software and martech in his VC and investment banking days. Feng was previously a Product Marketing Manager at Full Circle Insights. He is an avid proponent of expressing clear product value propositions, providing quality educational content, and being relentlessly analytical with your audience engagement.