Marketing attribution is the answer to a very simple question: if you could map out everything a person did that motivated them to become your customer, what would that journey look like?
Once we look at marketing attribution in this way, it begs the obvious follow-up: if you had a map like that for every customer of your business, what would you do with that information? The possibilities are almost endless, from lowering customer acquisition cost for your new mobile app, minimizing ad spend in certain geographic regions, and getting insights into new markets with even greater growth opportunities.
The rewards for answering these questions are clear, but getting there requires a deeper understanding of marketing attribution. This includes how advertising analytics compares to other types of analytics, different digital marketing attribution models, and what success in marketing attribution looks like for different types of companies.
While there’s no such thing as perfect attribution in marketing, the goal of every marketing attribution program should be to build a model that comes as close as possible to predicting the future.
Table of Contents
We’ve already hinted at what you could do with a map of your customers’ actions from awareness through conversion. That map (attribution) would make all of your marketing decisions much easier and less costly. Now, ask yourself:
Marketing attribution includes all the actions we can take to map out that customer journey. More specifically for this piece, attribution is the method by which growth marketers, demand generation marketers, and all the teams that support them justify their paid advertising programs. Marketing attribution - advertising analytics - is so important because it can show everyone in the company the most efficient ways to attract more of the right kind of customers.
There are many different kinds of analytics, and they can mean many different things depending on your audience. In this article we’re focused on advertising analytics, but you’ll also hear about web analytics, product analytics, and marketing analytics in general as methods for exploring your customer data.
There are a few primary goals for marketing attribution:
That last point is return on ad spend (ROAS), and it’s one of the bedrock criteria of an effective marketing attribution strategy.
There are many other things that you might want to know about your customer, including how much time they spent on your site or in your app, whether they successfully navigated your product, or what type of item they purchased from your site. While these are all valid and important questions, they can best be answered by some of the other types of analytics we mentioned above.
Ultimately, the goal of advertising analytics is to help you benchmark, test, and optimize conversion rates in your customer journey in order to grow your business. In essence, this comes down to four main goals:
Think of the customer journey map we mentioned above. The more accurately you can target every customer touchpoint in your marketing attribution efforts, the more clearly you can see a path forward to each of these four destinations.
Through your advertising efforts, your marketing attribution will allow you to optimize your channel performance and ROAS, so that every dollar you spend generates more customers at a lower cost. Digital marketing attribution can also help you generate more revenue more often and more revenue per customer, although these outcomes are secondary.
The tools you use to understand and achieve any of these four outcomes is critical to your success. Though as you explore which tools to rely on for your digital marketing attribution efforts, be aware of the different digital marketing attribution models at play.
There are two main approaches to counting things when building attribution models. Deterministic models count everything that happened during your customer’s journey from awareness to conversion. The alternative is probabilistic modeling, which uses statistics and/or machine learning to “guess” what happened. We’ll also touch on Media Mix Modeling as an alternative to a strict deterministic attribution model.
The deterministic approach is where most marketers should start if they haven’t worked with attribution models before. This is because all the customer data from each of their measurable touchpoints is counted in a deterministic model. This means it is inherently auditable, and that marketers can verify and deconstruct what has happened for each unique customer journey map.
A marketing attribution model allows you to assign proportional credit for a conversion to a channel or touchpoint in the customer journey. For example, if one of your potential customers clicks on one of your paid social ads, then they click on a ppc advertisement, and then they complete their purchase on your site, the model you choose will determine which channel should get “credit” for the conversion. The key is that the percentage that you allocate for a conversion must add up to 100%.
Consider which digital marketing attribution model is the right fit for your team as you set your marketing goals:
As we mentioned earlier, these models rely on statistical methods or machine learning to return an estimate of how much credit you should award to each of your marketing channels for every customer conversion. These probabilistic models are most commonly used to measure “incrementality” in campaigns, or to deduce meaningful signals in your data when you don’t have all the underlying deterministic information.
The best probabilistic insights come from solid deterministic data and a large sample size. This means you need lots of visitors to your site and lots of conversions to generate a probabilistic digital marketing attribution model with high confidence.
The potential payoff of probabilistic attribution models also reveals their greatest challenges. If you don’t have a large sample size, it’s unlikely your attribution predictions will be accurate. And the same holds true if your data isn’t reliable in the first place.
You can get a sense for whether a probabilistic model is likely to be correct by knowing its “p-value” or p-score. The p-value tells you how likely it is that the observed data would have occurred by random chance. The lower the p-value, the higher the likelihood that the output of the model is reliable. If you are paying someone to do probabilistic modeling for you, make sure they can tell you the p-value of their output.
As more and more people demand control over their online privacy, and governments all around the world respond with new regulations, ad platforms and marketers must adapt. Certain types of probabilistic marketing attribution like fingerprinting rely on making best guesses of a user’s identity, and these methods are becoming increasingly risky as privacy standards ramp up. Be aware of these methods when you construct your digital marketing attribution model, so you can balance the needs of your customers and your business goals.
While it’s beyond the scope of this article, some common probabilistic models used for digital marketing attribution include Shapely, Markov, and random forest.
Media Mix Modeling (MMM) uses a mathematical model (usually multiple regression) to estimate how marketing initiatives contribute to a desired outcome. These tests generally focus on one specific thing that matters to the business such as revenue, profit, or leads. MMM also often includes non-marketing variables (such as sales volume and pricing) and attempts to explain why different marketing initiatives have different levels of effectiveness.
Generally speaking, MMM complements deterministic, identity-resolution based marketing attribution models because it doesn’t rely on device matching. Media Mix Modeling can reveal details of a customer’s journey that a deterministic model might miss, particularly if mobile devices are involved.
There is no one-size-fits all “best” marketing attribution model for every business. Indeed, you might discover that different attribution models make sense at different points in the evolution of your business. The best marketing attribution model for your team depends on your goals and the type of business you run. The keys remain: accurate, consistent, reliable, deterministic data that is fully auditable.
Ultimately the model you pick is less important than everyone at your company knowing what it is and buying into it for the decision-making process.
Talk to one of our experts about your marketing attribution goals
If you work with a deterministic marketing attribution model, you might notice something unusual after a while: when you add up all the conversions attributed to ad platforms, and then match them against the actual conversions you have in your database, the ad platforms will always show more.
In order for ad platforms like Meta, Google, or LinkedIn to optimize your campaigns, they require you to send them lots of data about your business, including conversions and revenue. What the ad platform does with this data and how they use it is not totally transparent. There are few other instances where a company would send one of its biggest vendors some of its most sensitive information, in order to help the vendor determine what price they should charge for their product.
When someone clicks on an ad and visits your website, the ad platform can “see” the visit using a piece of JavaScript you have placed on the page for them, usually called a “pixel.” Then, if someone converts on your site within a certain window of time (the “Conversion Window”), the ad platform “sees” it and takes credit for that conversion in its reports. The Conversion Window is usually a 30 day period, which means that if someone clicks on an ad and then converts at any time within the following 30 days, that ad platform will take all of the credit for that conversion.
This “Conversion Window” (sometimes called the “Lookback Window”) can create more questions than it solves.
Imagine a situation where the Conversion Window was only one day. During that day, one of your customers clicked on ads from two different platforms in sequence (for example, Meta and Google) and then converted on your site. Both platforms can “see” that someone clicked on their ad, and they can see that same person converted during the Conversion Window. Both ad platforms take credit for the conversion, which can immediately introduce errors into your digital marketing attribution model.
In this simple example we’re only considering two ad platforms. Multiply this across a dozen ad platforms or more and you can see where many advertising analytics programs break down.
For simplicity, assume the user journey happens in the same day:
- Spent $5 for a click on Facebook
- Spent $5 for a click on Google
- User buys your product for $8
Facebook reporting says, “Great! You made $3 from our click!”
Google reporting says, “Great! You made $3 from our click!”
Math:
Cost = $5 + $5 = $10
Revenue = $8
Profit -$2
While Google Analytics (GA) is one of the most common attribution tools for marketers, it comes up short in several ways:
Google Analytics formerly used session tracking methodology, which, while helpful for determining what happens during individual online sessions, cannot tell you who took those actions. By sunsetting Universal Analytics (UA) on July 1st 2023 and forcing marketers to move over to GA4, Google embraced an event-based analytics methodology. This was intended to make it easier for marketers to map attribution throughout the complete customer journey, but GA4 has not offered much of an improvement over UA.
Marketers have realized that even though GA4 is easy to use, they still can’t get the answers they need about attribution. Those questions often require more detail than they can discern through GA4.
Identity resolution is required for accurate marketing attribution. Modern tools that leverage identity resolution are able to bind customer sessions together over a very long time period. Essentially, if tracked properly, the customer journey lasts forever. In addition, if you know “who” did something, it’s easier to segment users into useful cohorts for reporting and targeting. Finally, with identity resolution, virtually all your data is fully auditable. You can look at every interaction with a customer and compare it with known data sources to verify accuracy.
As hinted above, one promising way to connect different customer sessions together is to distill the customer journey down to individual events. If you can track your customers’ every interaction with your digital advertising, then you’re much closer to realizing the promise of marketing attribution.
Where GA4 has fallen short in the realm of event-based advertising analytics, other tools like Attribution App, Mixpanel, and Amplitude are stepping up. These solutions use an identity-based framework for measuring customer journeys, and they combine ease-of-use with the ability to get even deeper insights into your digital marketing attribution.
Just as there’s no one-size-fits all “best” marketing attribution model for every business, there’s no single attribution tool that will always return optimal results. The best fit for your team depends upon the state of your customer data, your goals, and the nature of your tech stack.
Calendly is a well-known product-led-growth company that needed a way to measure marketing’s impact on their sales pipeline. Lacking these insights, Calendly had to wait until sales registered a contract as, “Closed, Won,” one of the most common lagging indicators of success.
With help from Mammoth Growth, Calendly was able to unify campaign touchpoints such as webinar attendance, live event attendance, and off-site form fills from their Salesforce instance with revenue reporting in Attributionapp.com so they could accurately credit each touchpoint in a prospect's journey. Calendly now has the data required to make informed decisions and drive business growth forward. Instead of waiting months to see whether their marketing efforts would pay off, their marketing team can now get insights only days after launching a new campaign.
Reali is an end-to-end real estate platform that already had a strong tech stack in place when their growth began to accelerate. As they added more and more data sources, Reali discovered that they weren’t able to track a significant portion of their users. They turned to Mammoth Growth to organize their data infrastructure and improve their attribution model.
Mammoth Growth discovered a mismatch between data flowing from Reali’s customer data platform through the rest of their tech stack. To clear this up, Mammoth Growth built a custom attribution solution that would merge user data from multiple sources. With this updated marketing attribution model in place, Reali immediately had a much clearer picture of the full customer journey, and they were able to reduce their untracked users by 25%.
Barre3 offers unique exercise programs at their brick and mortar studios as well as through their online subscription service. While Barre3 planned to open more exercise studios, they were hampered by holes in their marketing attribution. Specifically, they found it difficult to map their customers’ journey across their exercise studios, their subscription service, and their ecommerce site.
Mammoth Growth teamed up with Barre3 to align their offline data with their customers’ actions online. With these updates to their marketing attribution plan, Barre3 could get insights into their customers’ in-studio behavior and its relation to their online behavior. For the first time, Barre3 had a 360-degree view of their customers’ journey. Finally, Mammoth Growth guided Barre3 through new solutions within Facebook and Google, allowing them to match their online ads with verified in-studio conversions.
By this point, it should be clear that accurate marketing attribution depends on many interdependent factors. It’s rare that a company has a gap in their attribution plan that can be perfectly filled by one specific piece of technology, with no changes to how they manage their data or interpret their customer journey.
In order to create a marketing attribution plan that’s aligned with your goals, it’s helpful to take a step back and look at the underlying architecture of your tech stack.In order to address companies’ far-reaching marketing attribution goals, Mammoth Growth developed our Analytics Architecture program. The objectives of this 8-10 week project are a complete audit of a company’s tech stack & data infrastructure, and a roadmap for improvements to their marketing attribution plan. Our Analytics Architecture program allows Mammoth Growth to develop a deeper understanding of a company’s existing marketing attribution methods and pain points.
When approaching digital marketing attribution projects, Mammoth Growth follows these steps within our Analytics Architecture project:
For every company, there’s a unique combination of marketing attribution model + attribution technology + process improvements that can increase ROAS while delivering more revenue from more customers. Finding that sweet spot isn’t a matter of luck. It takes discipline and focus to find the right fit for your team. Contact one of our experts today and let’s talk about your digital marketing attribution goals.
