Imagine yourself buying a new dog bowl versus shopping for a couch. The customer journey for one is simple, while the other is much more complex.
Shopping for a new dog bowl might be a 2-minute exercise with minimal research—a dog bowl is after all significantly cheaper than a couch. You could look around and quickly buy something that looks cute or has the features you want.
A couch, on the other hand, is an expensive purchase that could take weeks of consideration. This involves checking multiple sites to find the best deal, having to go back and forth to discuss with the people you share your home with, and possibly taking some in-store visits to try out the merchandise. All the while, you’re seeing ad after ad trying to draw you back to a website to purchase. It is a layered process, with many different touch points at play simultaneously.
It is obvious that these customer journeys are very different—and their attribution models would accordingly be different as well.
If you’re selling dog bowls, you could use solely last-click attribution or maybe a mix with post-purchase survey data. This would suffice in getting you a good picture of what marketing touch points mattered the most.
This is because your customers likely don’t need many touch points before making a purchase. It might be an immediate impulse buy through social media adverts, or purchasing the first result you get in a Google search.
As for couch buying, it has many more touch points as a result of the layered customer journey.
Your business could be doing podcasts, billboard and bus stop advertisements, digital campaigns, and so much more. All this would play a role in your marketing strategy, by slowly nudging the customer closer and closer to purchase.
With a complex mix to meet a complex customer journey, attribution is a lot harder here, and requires a different approach in order to capture the return on investment for each touch.
We use dog bowls versus couches as an example here, but they represent the broad spectrum of customer journeys that exist. For your brand, your products could also be simple or complex. By identifying that and hence understanding your customer’s journey, we can pick an attribution model that will benefit your business.
Before we dive further into attribution modeling, let’s first elaborate on the data and data sources. Without this data, our models would be empty and make all modeling efforts useless.
Post-purchase surveys
Post-purchase surveys (PPS) are one or more questions shown to a customer immediately after they’ve bought a product on your site.
They are a reliable source of data. It is largely due to the nature of the data collected, being zero-party data. This means that the data is accurate and not subject to legal privacy constraints.
PPSs enable you to find out which touch point made the most lasting impression on the user. If you do a lot of advertising (think awareness or branding) that doesn’t require a click to your website, this will fill in a lot of gaps left by GA-based models.
However, it is not perfect. The main reason here is that customers may misremember or over-emphasize on a specific touch point. Although, there is an argument to be made that the touch point that sticks out the most in their head may be the most important.
Only a fraction of customers will supply this information given the nature of surveys and their typically lower response rates (RR)—although our friends at Fairing have figured out a way to regularly get 30-40% response rates with their Question Stream.
Another drawback: you will also get info on how the customer heard about you, but not any other touch points. This means that the scope of your data is tight.
Discount codes
Discount codes are alphanumeric strings that your store can offer to encourage purchases. They are usually associated with an overarching promotional marketing strategy.
They are a reliable and effective way to track your marketing efforts. They let you see which channels are generating the most traffic, conversions, and so on. Whenever a customer keys in the code, your store can track where the discount code brought them from. It gives you visibility into its effectiveness that GA-based models can’t compare to.
Discount codes are most beneficial if your brand is heavy into podcast advertising, given the difficulty in tracking a podcast impression all the way to a session on your store. This means that it only gives insight into channels like podcasting that rely on discount codes for attribution. If this does not suit your marketing mix, then discount codes will likely not be of much value.
Google Analytics
Google Analytics (GA) is a platform that collects data from your websites and apps. It aids in creating reports that provide insights into your business. It gives you a full history of online interactions from the customer.
Not only that, the platform also enables you to look at attribution through different perspectives: first click, last click, last ad click, and more.
However, this is the least reliable of the lot. It functions by stitching together touch points from the same user. This relies on cookies, which is fragile for users who use other browsers like Safari. This is because of things like cookie expiration and the fact that a user may be browsing on multiple devices before purchasing.
GA also doesn't account for users shopping from multiple devices. An example is if a customer visits your eCommerce store from their work laptop and then from their phone. GA will recognize that person as 2 different users.
Additionally, the presence of ad blockers will cut out a portion of this info—rendering your data inaccurate or incomplete. You also do not get any visibility into touch points that do not include a visit to your website. For brands that advertise a lot on offline channels (such as OOH, podcasts, subways ads, etc), GA is not enough to get a full picture.
What attribution models are available?
Now that we’ve covered the data sources available to you, let’s talk attribution models.
What’s important in modeling is choosing one that would suit your business. This is why the dog bowl and couch companies would have to choose different attribution models.
Basic single-source attribution models
In the realm of attribution models, survey-based and discount-code attribution are the most basic and direct forms. Here, you simply account for the data you collected from each source as the name suggests, and analyze from there.
While these models have their perks in terms of simplicity, they’re dated and would not suffice in the modern digital world.
Survey-based attribution
This looks only at the customer’s response to your PPS survey, specifically to the question of “How did you hear about us?”. For the strengths and weaknesses of this model, refer to the evaluation of PPS in the section above.
Discount-code attribution
This assigns credit for a purchase to different channels and vendors based on the discount code value associated with an order.
Attribution models for simple customer journeys
Here, GA-based attribution models are catered to simpler customer journeys, much like consumers of the dog bowl. They are able to aid businesses whose customers have fewer touch points, and a relatively simple funnel towards purchase.
GA-based attribution models
These are attribution models based on Google Analytics tracking. Under this umbrella, there are numerous options available (also available through Daasity). Most of these models share the same drawbacks: you’re assigning credit to a single touchpoint, and these models don’t account for offline sources that aren’t picked up by GA4.
The first is Last Click. This model attributes credit to the traffic source that initiated the session in which a user made a purchase.
- Pro: Relatively reliable because it is based on the most recent marketing touch point trackable through your site
- Con: You are choosing a single touch point to assign credit
Following that is First Click. It attributes credit to the first traffic source the user interacted with in the 30 days prior to the purchase.
- Pro: Gives you insight into upper-funnel activity
- Con: You are choosing a single touch point to assign credit
Next is Assisted. This assigns credit for an order to every non-last-click traffic source that the user interacted with in the 30 days prior to purchase. Think of this as "non-last-click" attribution.
- Pro: Gives you insight into upper-funnel activity without choosing only 1 single touch point
Then we have Last Click + Assisted, combining the Last Click and Assisted attribution models. Think of this as "any click" attribution. This is because a traffic source will get credit for the order if there was a touch point at any point in the 30 days prior to purchase.
- Pro: Gives you insight into how many orders / how much revenue were influenced by a particular channel or vendor. It doesn’t arbitrarily choose a single touch point, and doesn’t rely on a data-driven model that could be biased
- Cons: The general pitfalls of GA-based models
Last Ad Click. It attributes credit to the last advertising channel the user interacted with before making a purchase.
- Pro: Reclaims credit for paid touch points that may get overshadowed by other organic touch points
- Cons: You are choosing a single touch point to assign credit
Lastly, we have Last Marketing Click. This attributes credit to the last marketing channel the user interacted with before purchasing.
- Pro: Reclaims credit for any marketing touch points that may get overshadowed by other organic touch points
- Cons: You are choosing a single touch point to assign credit
Back to our couch buying journey:
Using discount code attribution would hence be very important here if that were a part of your company’s strategy—if your brand advertises heavily on podcasts, for example.
PPSs would also be crucial to understanding the complex buying journeys because your data is likely only capturing a small part of that journey. The zero-party data can fill you in on the different contexts that the other data cannot provide.
When zero-party data isn’t available or sufficient, you’ll likely also want to look at an attribution model like Assisted + Last Click. This will give you full visibility into all the digital touch points that led up to a purchase.
Assisted + Last Click models attribute the sale not just to the recorded last click’s source, but to every touchpoint along the way. This stops brands from placing too much emphasis on the last touch (or any specific touchpoint).
Where this doesn’t quite cut it, there’s another option: a custom attribution model built specifically for your brand’s needs.
Attribution model for complex customer journeys
A custom attribution model uses a waterfall approach to assign credit to each order based on the most reliable attribution info available. If there is discount code attribution available, it will use that. Otherwise, it will look for post-purchase survey data and use that if available. If there’s no post-purchase survey data available, then it will use the GA-based model of your choosing.
Utilizing such a model would give you great flexibility, and also the most accurate data.
Go with what’s best for your business
A bird in hand is worth two in the bush—not every business is ready for the complexities of a custom attribution model. In the days before digital marketing, entire conglomerates were built on coupon codes and surveys. But if you’re ready to incorporate multiple sources of data into your model, Daasity can help.
