Attribution can be tricky for online sellers. The customer of today is multichannel and multi-device, with an average of 5.5 touchpoints between discovery and purchase.
Figuring out how to credit each channel with a sale can be a real challenge, especially if you aspire to use a multichannel marketing attribution model.
Many online sellers turn to position based attribution models, relying on the position of each touchpoint to determine credit for a sale.
However, if you’ve ever read one of our attribution articles, you know that position based attribution models are inherently flawed from their very conception.
This flaw has a butterfly effect, as we’ve seen at Divvit that 14% of all sales are misattributed.
But where do we go from position based attribution models? If they cause flaws in how we attribute, we need to think smarter. Smarter than humans even.
This is where data driven attribution models come in. Using Machine Learning algorithms, they can assign true credit to a channel based on predictive analytics and customer behavior.
In this article, we’ll compare and contrast position based attribution with data driven attribution and make the case for why value needs to be at the core of your attribution strategy.
Position Based Attribution Models: What You Need to Know
We’ve compared and contrasted different marketing attribution models on this blog before. There are many categories of attribution models, but we already know that multi touch marketing attribution models provide a better representation for multichannel customers.
And today, all customers are multichannel customers.
However, not all multi-channel attribution models are created equally. Linear, position based, time decay, these models all carry the same pros and cons by virtue of the fact that they’re based on the position of the touchpoint in relation to conversion.
The biggest benefit to position based attribution models is that it provides a simplicity in an otherwise complex subject. Even time decay attribution, which is the most complex of the three, tends to be rather easy to understand.
Certain things should be kept simple, especially when we talk about design and UX.
You know what shouldn’t be simple?
Your marketing attribution model.
Why? Because your customers aren’t simple. Their journeys aren’t simple. And though your sales funnel might be as streamlined and simple as possible, your customers’ journeys are going to be individual and unique for each and every customer.
The very fact that position based marketing attribution models are based on position is their greatest flaw. A touchpoint’s position plays such a small role in the grand scheme of whether or not a channel provides value for your online store.
To focus only on the position does a disservice to you and your marketing budget. Often, more emphasis is placed on leading and converting channels and assisting channels are well undervalued, when they could have been the real deciding factor for the customer’s purchase decision.
Some models are better than others, but they all have that same fatal flaw.
So what do you use if position based attribution models are out?
This is where data driven algorithmic attribution models come in.
Data Driven Attribution: Using Algorithmic Attribution for Smarter Insights
The biggest difference between position based marketing attribution and data driven marketing attribution is that position isn’t the only factor taken into account.
At Divvit, we count 5.5 touchpoints across 2 or more devices between the initial discovery and the purchase. Each of these touchpoints is an opportunity to move the customer along towards conversion.
So why would you use a position based attribution model that disregards the value in each an every one of these touchpoints?
There is so much more that goes into determining whether or not a channel provides value:
- What pages did the customer consult?
- How long did they spend on your site?
- Did they add things to their cart?
- Did they begin a checkout process?
These other factors are critical to deciding whether or not a touchpoint in your customer’s journey helps factor into the purchase decision process.
Data driven attribution models typically employ an algorithmic attribution model. These algorithms use Machine Learning to determine the value in each individual touchpoint and assign a score of how valuable it is.
Sometimes, when we think of words like “algorithm” and “Machine Learning,” we think “complicated.”
Complicated isn’t quite the right word. Complex is a much better fit, and complex doesn’t have to be complicated.
How Data Driven Marketing Attribution Uses Machine Learning Algorithms
Data driven marketing attribution uses Machine Learning to assign a value to each touchpoint based on the behaviors your customer exhibits during their session on your site.
This Machine Learning algorithm learns from the behavior as your customers interact with your site. Meaning that if many customers converted after having done a certain sequence of actions, this algorithm would assign a probability of conversion based on the sequence performed by a new customer.
Based on the touchpoint used to come to the site, the algorithm assigns a probability of conversion and a percentage of credit to be attributed. This is based both off of the probability that the channel usually has for your customers, and the sequence of behavior your customer performs.
That behavior makes a huge impact on the credit attributed- for example if a customer just looks at one or two products, it might have a lower conversion probability. If the customer adds things to their cart or looks at the shipping and returns policy, it might have a higher conversion probability.
The point is, the algorithm gets better and more customized as it processes more data. The more data it has, the more it will be able to predict conversion with your specific customers.
This might seem terribly complicated to use, but it’s actually quite simple. Once the tracking code is implemented on each page you want to track, it’s as easy as looking at your analytics.
Why doesn’t everyone do this?
A lot of people shy away from data driven algorithmic attribution models because they have the perception that they’re difficult to use and expensive to implement.
The truth is that they may be complex to create (writing algorithms isn’t exactly my own forté), but using one that’s already created can be quite simple. As far as expensive, it could get pricey to hire your own developer to code and create your own data driven attribution model.
But why reinvent the wheel?
There are already data driven algorithms you can use to achieve this same true attribution for your online store. They don’t have to be expensive and can be used by a variety of ecommerce stores.
Marketing attribution seems like a complicated topic until you break it down. When you understand the impact that it has behind your ROI and how you can use it to make better decisions for your online store, it can make a difference in how you use your marketing budget.
Using data driven attribution can help take a lot of the guesswork out of the process while attributing based on the real value each touchpoint has.
While position based attribution models may be simple to use, the simplicity can cause you to lose money by over or undervaluing certain channels.
After all, is the simplicity worth 14% of misattributed channels?
What attribution model do you use? Have you ever used data driven or machine learning attribution?
Tell us below or tweet us!