If you don’t have a clear idea about which attribution models are responsible for conversions — and which ones are not — it’s hard to measure the success of an individual campaign or channel. Furthermore, confidently committing a budget to new campaigns can be challenging.
For a long time, marketing teams have used attribution models to measure the impact of particular channels. For example, modeling could tell if PPC ads, SEM, or video marketing drove leads and sales.
However, these models are far from perfect. Sadly, they often produce inaccurate results, resulting in marketing teams plowing budgets into underperforming channels or underfunding marketing that genuinely had an impact.
So, let’s look at a few popular models, talk about the problems with attribution, and get a glimpse of how attribution can evolve as third-party cookies are gradually phased out.
Popular attribution models
Before examining their faults, let’s examine a few of the most popular attribution models today. Perhaps the easiest way to do this is to think about a customer.
Let’s say you run a store that sells footwear. For simplicity, we’ll say your customer buys some shoes after interacting with four marketing touchpoints.
They are a:
- YouTube ad
- Display ad
- Social media post
- PPC Search ad.
The customer sees each ad in that order, pressing buy soon after seeing the PPC Search ad.
In the scenario described above, let’s look at a table explaining how attribution is credited to each channel (with a percentage total).
*Time-Decay % varies depending on the time between touchpoints.
As you can see, attribution doesn’t always represent the impact of a campaign.
For a deeper dive into the subject, read our article on 6 attribution models you should know about.
Problems with current attribution models
There are a few major problems with current attribution models. Some are difficult to solve; others can be overcome using more sophisticated methods.
1. They oversimplify reality
The moments that influence human behavior are extremely complicated. Digging deep into human motivation is a problem better suited to philosophers and neuroscientists.
By design, attribution modeling is simple. It takes messy data and imposes a narrative over the top. Quick and dirty calculations have merit, but they can’t always get to the truth. This means you could easily be over or under-attributing a particular channel or channel’s effect.
2. Attributed ROI is misleading
Another attribution issue is that it struggles to capture incremental ROI. If your brand has been operating for a while, you’ve probably built goodwill, contributing to sales via word of mouth. Attribution modeling can’t capture this.
3. It doesn’t measure dark social
Dark social consists of all the conversations about your product that can’t be measured or tracked by attribution tools. Think emails, private messaging, and personal conversations.
Imagine this scenario. You spend months researching a product, watching webinars, reading content, etc. You tell your boss about the product via text message. They go directly to the site and purchase it.
All the content you have consumed won’t be counted despite its huge role in your decision.
4. Many models ignore the customer journey
Research suggests that each sale has about six to eight touchpoints. Many models only count one of these touchpoints and give it all the credit. However, what we know about the customer journey is quite different.
5. The role of external factors
Pricing plays a huge role in whether people buy products. How a business price an item relative to its competitors greatly affects eventual sales.
A customer who is researching a product might look at a few options. Let’s say yours is cheaper, so they buy that. You might attribute it to SEO marketing when it was the price point that actually did the work.
Afterward, you could be inclined to invest more in SEO when you should think of aggressive pricing against your competitors.
Are attribution models all bad?
Look, we’re not saying all attribution models are terrible. They have their utility. However, you want the most reliable data to measure your campaigns, especially when using that information to inform your marketing spending. So, looking at it that way, you need something better.
Most attribution models rely on third-party cookies. As they are gradually phased out, these models lose some of their power. This brings us to another thing to consider: Google Analytics 4 (GA4) and the changes it will bring.
GA4 Migration
After a long wait, Google Analytics 4 is here. While Universal Analytics (UA) will stick around for a while, it is being sunsetted. So, if you’re serious about attribution, you’ll need to migrate to GA4.
There are a lot of new things to consider about GA4. For starters, how does it handle cookies? The bad news for marketing teams is that it restricts third-party cookies. However, it does use first-party cookies.
Another thing to note is that GA4 incorporates some other signals into the mix. Namely, session data that will capture both app and browser sessions. Apps and mobile browsing are taking over. In 2021, a staggering 90% of smartphone time was spent on apps. Combine this with the cookie’s end, and it becomes clear why UA is being sidelined.
Transitioning to GA4 will be a challenge for some. Many teams have set up workflows with enhancements and functions for UA. Many of these won’t translate to GA4. Additionally, UA goal metrics will need to be converted. For example, Signups in UA will now be conversions in GA4.
However, some of the upsides of GA4 are:
- Better analytics
- Better reporting
- Automated insights
- Event-based tracking
Why GA4 is better for attribution
UA has many of the problems associated with attribution. In particular, using last-click attribution is an issue for teams who want to credit marketing channels accurately.
One problem with last-click is that it fails to account for other touchpoints and isn’t representative of the buyer journey. GA4 takes a more data-driven approach to attribution, similar to Amanda AI. Google has employed machine learning algorithms that will combine with users who have signed up for “Ad Personalization.” Overall, the results should be far more accurate.
Amanda AI uses an impact-driven model. It works with Google’s algorithms to ensure your budget is allocated to ad types and channels, making a difference. We welcome a shift from the attribution models listed above because of their inaccuracy. We want our customers to have the best results, so Amanda AI gathers data from one place, Google Analytics, and optimize the campaigns based on audiences with the highest conversion probability. That’s impact-driven to us.
