People always ask us which attribution model is THE model a prospect or customer should choose. I know people don’t like it when consultants answer, “It depends,” to every question, but people dabbling in marketing attribution need to understand that they will need more than one model because different models answer different questions.
To choose the right model, you must understand which question you, your boss, or your company are trying to answer. For example, do they want to know which campaign performs best at a particular stage so they can optimize for that stage? Or do they want to represent how much pipeline or bookings marketing contributes to the business compared to other departments?
You’ve probably guessed (correctly) that each of those questions is tied to a very different model.
Instead of waxing poetic about a single model, we’ll give examples of when a specific model makes sense and when it doesn’t. But first, we need to talk about attribution as a philosophy.
What is Marketing Attribution?
The original intent of attribution was to understand which marketing tactics are most effective at generating pipeline and bookings for a business (i.e., campaign optimization). However, most business leaders use attribution to indicate how much “credit” a department should get for generating opportunities, pipeline, and bookings versus another.
When businesses move away from using attribution as a campaign optimization tool and into the land of proving marketing’s “worth,” it can get messy fast. But because marketing leaders are under tremendous pressure to prove that what they’re doing is working, we realize that the underlying motivation behind implementing attribution isn’t going to change any time soon in B2B.
To do attribution well, you need to have proper CRM usage by the sales team. They need to use their CRM as a forecasting tool. You’ll also need a marketing automation tool integrated with your CRM. Ideally, you’ll also have UTM tracking methods securely in place and use the campaign and member objects in your CRM. Read our introductory article on marketing attribution if you are looking for more basics on the topic.
Why Do People Use Single-Touch Attribution?
Most people use attribution without even knowing it. “Opportunity Source,” “Primary Campaign,” or “Lead Source” are all single-touch attribution models. These models assign 100% of the opportunity dollar value to whichever tactic or department interacted with the prospect at that point of measurement.
What Is a First-Touch Model?
A first-touch attribution model assigns 100% of an opportunity’s value to the prospect’s first brand interaction. It’s also known as the “Lead Source.” It answers, “What are the best tactics for getting someone to engage with our brand first?”
In Salesforce, this is how the “Primary Contact” on the opportunity was created in your database. The lead source could include list purchases or true interactions with marketing and sales. And it doesn’t matter how long ago this touch happened in Salesforce.
In CaliberMind, we use the first touch that happened against anyone on the account (including leads), and we generally look back 365 days from opportunity creation. We recommend that you look at how long it typically takes sales to close an opportunity after opening it and then use that as your starting point to figure out how many days before opportunity creation is a reasonable time frame for a “first” touch.
With a first-touch model, we’re looking for the first signal that the company is researching possible solutions for a problem. Generally, that will be 50%-100% of your deal cycle lifespan. For example, if your sales team closes a deal six months after opening it, you should plan on looking back 90-180 days from opportunity creation.
What is a Last-Touch Model?
A last-touch model (also known as the middle-touch model by some attribution companies) assigns 100% of any opportunity’s value to the interaction immediately before opportunity creation. It’s also known as “Opportunity Source” and answers, “What is the best channel to get our prospect to engage with sales?”
In Salesforce, Opportunity Source is the primary campaign associated with the opportunity. This requires sales to create the opportunity on the “correct” contact. It’s also dependent on how you configure your system. If you want to capture partner-sourced opportunities, you’ll need to leverage campaigns. The same is true for inside sales or account executive-sourced opportunities.
In CaliberMind, by default, we look across the account (including leads) and identify the last inbound touch. An inbound touch is a proactive interaction by your prospect, NOT outbound attempts by your company. We also create what’s known as “virtual” campaigns to capture interactions with channel, product, and sales by looking at tasks, events, web signals, form fills, and other activities. Looking at multiple signals makes it possible for other departments to get “credit” for sourcing an opportunity without requiring more data entry from sales.
Is There More Than One Last-Touch Model?
Yes. The “other” last touch model looks at the touch just before sales closes the opportunity as opposed to the touch just before the opportunity is opened. It answers, “What is the one touch that inspires the prospect to commit to purchasing?”
Since the last touch is usually hot on the heels of a meeting with sales, this is rarely used in non-PLG (Product-Led Growth) organizations.
Why Do People Use Multi-Touch Attribution?
MTA is used to answer the question, “How much pipeline does each department produce for the business?” but it was intended to help demonstrate the broader value marketers bring to the organization. Multi-touch attribution (MTA) was created because B2B buyer journeys are complex. Purchases involve multiple people at different points of the decision-making process, and it’s not uncommon for priorities to change and for people to disengage or re-engage.
This brings to mind one of our favorite buyer journey diagrams:
Instead of focusing on a single point in time, MTA also allows marketers to show when campaigns that aren’t good at lead generation are good at keeping an in-flight opportunity engaged. All MTA models can answer, “How much pipeline or bookings does each team contribute,” but some are better than others at answering specific, detailed questions.
What is an Even-Weighted or Linear Attribution Model?
Even models assign all touch points an equal portion of opportunity dollars. Typically, this model looks back 365 days before opportunity creation until opportunity close, so the attribution dollars aren’t finished being divided until an opportunity is closed and touch points are no longer being “credited.” Even models remove the assumption that a single touch point or person on the account is more meaningful than others. Like most MTA models, it will naturally show more “credit” to the most popular tactics.
What is a W-Shaped Attribution Model?
The w-shaped model has some variation in its definition between attribution vendors. At CaliberMind, we assign more weight to the first touch, last touch, and touches associated with the primary contact. This model answers, “Which department or tactics are the most effective at engaging an account and hooking the primary contact?” This model is good for organizations that recognize most touches after the opportunity creation will be sales driven. However, it will not help marketers understand which materials and campaigns buyers engage with while working with sales.
Before the opportunity is opened, all other touch points still receive some dollar value, but it’s significantly less than the other categories. Zero dollars are assigned to touch points that happen after opportunity creation. This model works best if salespeople must create an opportunity from a contact, which you can force by removing opportunity-create options at the account level in your CRM.
What is a U-Shaped Attribution Model?
The u-shaped model assigns more credit to the first and last touch before opportunity creation. All other touchpoints before opportunity creation are also given a small amount of “credit” or opportunity dollars, but anything that takes place after opportunity creation is ignored. This model answers, “Which department or tactics are the most effective at engaging an account and getting them to engage with sales?” This model is suitable for organizations that recognize most touches after the opportunity creation will be sales driven. However, it will not help marketers understand which materials and campaigns buyers engage with while working with sales.
What is a Chain-Based Attribution Model?
The Markov effect or chain-based modeling is one of the most common machine learning models on the market. These models are meant to remove the bias that a single point in time or a single person is what makes a deal winnable or not. Instead, it answers, “What is the blueprint for a successful buyer journey?”
At CaliberMind, we look at touchpoints up to 365 days before opportunity creation until opportunity close. This is done by isolating buyers’ most common sequence of steps before purchasing the product. So we’re training the model against the success criteria of a closed-won opportunity. Then points are deducted when a prospect strays from this path. How many points are deducted depends on how much the deviation will negatively impact the odds of winning the deal.
A word of caution: Machine-learning models need many closed-won deals to accurately “learn” which sequence of interactions is the best. If you have less than 100 closed won opportunities, using a different multi-touch model is best.
Single-Touch vs. Multi-Touch Attribution
What About Custom Marketing Attribution Models?
Are there different variations of models? Absolutely. But there are a few reasons we see internally architected models fail.
- They’re expensive. A good attribution model requires connections to your marketing tools and your CRM. And it should even incorporate web signals. Each of these things is time-consuming to set up. The most expensive part of managing your model is change management. Marketers change technology a lot, and each time something changes, you have to map the data and update your model.
- They’re easy to over-engineer. All models have excellent intentions behind them, but none are perfect. We can’t track things like word of mouth, and digital tracking has a lot of gaps. We must acknowledge that any model is an estimate and be cautious of over-valuing specific personas and touchpoints. Weighing channels more heavily when they’re expensive may not reflect what works to attract people to your brand. Human bias is real.
- They forget WHY. Many models are over-engineered because the marketing department has difficulty selling the model to the rest of the organization. People who try to compensate by introducing complex weights or calculations usually have lost sight of their goal. If your goal is to prove that marketing “works,” ignoring sales and channel touches is often the culprit.
If you hate attribution, it’s time to ask yourself some hard questions. What are you trying to do with the model? Which questions are you trying to answer? And do I need help or a different tool to get attribution right? Hopefully, this article gives you some ideas on where to start looking.