A huge responsibility for marketing teams is knowing which campaigns are most effective—not only does this clarity inform you where you should focus your resources to acquire new customers, but it also allows you to reasonably plan and project future growth.
For the better half of the 1900s, marketing and advertising lacked the measurability that allowed campaign managers to effectively determine ROI and spawned this much-repeated quote:
"Half the money I spend on advertising is wasted; the trouble is I don't know which half." —John Wanamaker, early 1900s”
Fortunately, digital marketing today provides a more data-driven approach. We no longer need to "spray and pray," (or spray all types of content across all types of channels and pray that some of it performs.) Many tools out there have impression and conversion costs baked in, making it easy to determine ROI on various campaigns. And the tech community has ignited an irreversible trend of using data rigorously to drive growth.
But when you're running multiple campaigns in today's multi-device world, it can be tricky to measure ROI for each. How much influence did the Twitter ad impression make in this sign up? How about the branded Adwords campaign? If most of your conversions come from paid search, should you still continue to invest in paid social?
In this lesson, we'll lay out the fundamentals of marketing attribution. We'll look at pros and cons of the common attribution models, provide some guidance in selecting a model, and give some thoughts to building your own.
To better understand how different attribution models influence how key events receive credit, let's pretend that you have- just won an Oscar. Each attribution model presented below will define your approach in your "Thank-You" speech.
Last touch is extremely common in most analytics tools (like AdWords) and many programmatic ad platforms. Basically, the last campaign or impression that the lead interacted with gets all of the conversion credit. If you're only running one or two campaigns, the simplicity and out-of-the-box accessibility of this model is appealing.
However, in cases where you have multiple marketing channels, this sort of attribution is misleading and won't help you better allocate your marketing spend.
When you receive your Oscar, last touch attribution means you would only thank the last person who helped you win the award. Nobody else would get credit, and your parents may be a bit upset.
In addition to likely being overly simplistic, last touch attribution modeling also sways your results in favor of particular advertising platforms. It's not surprising Google emphasizes last touch attribution because once users are ready to buy, they will often search for your product and convert on that search, even if they've seen and clicked through many other display ads. Knowing this, you may also want to consider what the advertising platform's bias is in showing you last touch attribution results.
First touch attribution is the exact opposite of the last touch model, the first impression that the lead interacted with gets all of the conversion credit. Similarly, this is offered in most analytics tools, making it simple and easy to get started. Again, in cases where you have multiple marketing channels, this model is incorrect and won't give you the granular info needed to allocate future marketing spend.
In the Oscar analogy, you would thank the first person—your mom—that helped motivate you to get into your profession. No one else would get a mention, which means you may have shot yourself in the foot if you want to make movies again.
First and last touch attribution models, while flawed, are simple to understand. But the devil is in the details, especially when your customer journey spans more than one device or campaign. That's where multi-touch attribution comes in.
Multi-touch attribution is the act of determining the value of each customer touch point leading to a conversion. This helps you figure out which marketing channel(s) or campaign(s) should be credited with the conversion, with the ultimate intention of allocating future spend to acquire new customers effectively.
You can think of multi-touch attribution as a set of rules that gives variable credit or "weight" to different marketing channels. Or, more mathematically, it can be considered as an equation where one side has the customer's touch points as cost per impression and its unique weight; on the other end should be the conversion value (e.g., the value of a
Sign Up for your business).
For the Oscars, this means that you would thank multiple people for your award—your parents, your agent, your director, etc. However, what you say about each person you are thanking (giving them a tiny shoutout vs providing a more extended tribute) equals the weight of the credit you're giving them. In the below examples, we'll explore how you allocate the thank-yous.
The linear model gives all interactions the same credit in the lead's conversion. This is one step more sophisticated than first and last touch attribution and one step less wrong.
When giving out your Oscar thank-yous, everyone you mention will get equal airtime and, therefore, equal credit. You may even say this in a robot voice to be sure no single shoutout is getting special treatment.
Time decay attribution is a model that will give more conversion credit to interactions that happen closer to the conversion event. This makes common sense. Well-known analytics expert, Avinash Kaushik, advocates for this model, saying that the earlier touch points are weighted less, because "if [those] were magnificent, why did they not convert?"
During your Oscar thank-yous, you'd name the people who led to your success linearly, in chronological order. However, towards the end of your thank-you speech, each individual you mention will receive a longer, more deliberate recognition. The audience would understand that those you mentioned towards the end of your speech contributed more to your success.
This is more advanced, but you get to set your own weights. Many companies do a U-shaped model where they give first and last the most weight and then split the rest among every step in between.
In the next section, we'll give you some ideas and things to think about to help speed up the time it takes to create an effective custom weighted attribution model for your business.
For your Oscar thank-yous, you'd give varying degrees of credit to those who supported you. Maybe you'd thank your director most since she took the biggest chance on you and then your dad next for finally accepting the fact that you won't be a doctor.
Attempting to build your own model may lead to additional unnecessary complexity, but there are some situations where a custom model can help you better measure ROI and allocate future spend:
The key here is to take an existing weighted model, and then adjust it incrementally to see if it fits better with your business. It's also critical that the end conversions of these adjustments are carefully monitored. For instance, if conversion quality declines, then you should course correct quickly.
Here are some guidelines to help you figure out what to adjust:
Another way to think of this is what user behavior do you value more than others?
A micro conversion is a pre-conversion event that signals some value to your business. For example, a user who downloads a white paper might suggest more readiness to talk to sales, so this interaction may be given more weight than reading a top of funnel blog post. As a result, you can focus more of your marketing spend and effort to optimize for users that download white papers.
Again, keeping an eye on the quality of the final conversions is paramount. Focusing all of your effort and spend to funneling your users towards these micro-conversions may ultimately lead to targeting the wrong audience or incorrectly qualifying them for the conversion event.
This is a bit more involved and will take longer (depending on how many data points you have, e.g., how many potential customers are interacting with your brand. It requires experimentation with marketing spend in different channels and tracking the quality of the conversions down the funnel. This approach is also best suited when testing a new channel, e.g., a new paid social channel or a new ad network that can target a new audience.
The idea here is to use cohort analysis to measure the conversion values of a cohort when there is no spend vs. when there is spend on a channel.
Since the above yielded similar conversion rates (and, to be sure, we'd have to look at the eventual LTV of the conversion), then the first touchpoint of seeing a display ad has a negligible impact on the conversion.
Here are three important things to consider when selecting an attribution model.
What is more valuable to you: a product signup or a sales qualified lead? What is the average LTV of your customer? Do your customers pay in monthly subscription payments or one-time charges?
All of these questions guide your customer acquisition strategy, from identifying what channels to use to launch these campaigns to budgeting ad spend.
Multi-touch attribution is part art and part science (at least for now—and web technology has enabled tremendous progress since the early 1900s). And like most analysis, looking at trends or comparing between campaigns will suffice in terms of determining the strategies to double down on or the strategies to cut.
Moreover, the custom weights on a multi-channel attribution model will evolve over time, as the target audience evolves, your messaging adjusts, new channels are added, and your paid marketing budget grows. Many times it isn't worth the effort to make sure the model is perfect if it's going to change in three months. Getting it mostly the way there can often suffice in an environment that is changing quickly.
Ultimately, deciding the level of sophistication of your multi-channel attribution comes down to what your team wants and your available resources.
Last touch is easiest to implement and understand. But it discounts many other forms of advertising that are more subtle, like display ads or guest content on other blogs.
Certainly, the more complex custom models will take more time and effort. But depending on your monthly paid marketing spend, that additional effort may not be worth the residual marginal ROAS (return on ad spend) of getting the model near perfect. Plus, as you add new channels and as your marketing spend grows, so too will the weights.
It's sometimes best to just be directionally correct, forget about the last 10 percent, and spend your resources elsewhere. At least you'll know that 90 percent of your marketing is working!
Have you decided to use multi-touch attribution? What tools and model are you using and why? We'd love to know—let us know on Twitter!