Attribution models are essential tools in marketing, helping to determine which efforts are driving conversions and how successful they are. Most commonly used attribution nowadays can be put under the direct attribution category. This comes down to tradition, influence of the tools we use and necessity. But there is another approach to attribution that might be even easier to implement, not impacted by privacy changes and provide additional benefits – incrementality.
This article explores direct attribution and incremental attribution, highlighting their mechanisms, applications, and differences.
What is Direct Attribution?
Direct attribution is an attribution model that credits conversions to a source based on the interaction (view, open, click) with a marketing asset (ad, email, social media post). This is typically implemented by using tracking links, pixels, scripts, and cookies that collect information about the user and their interactions, which are then passed to the tool of our choice.
Examples of Direct Attribution:
- Tracking conversions of users who clicked through a Google Search Ad.
- Tracking conversions of users who opened or clicked an email.
- Tracking conversions from a social media post that used a link with UTM parameters.
What is Incremental Attribution?
Incremental attribution is an attribution model that measures the additional impact of marketing efforts beyond what would have occurred without them. This approach involves comparing a control group, which is not exposed to the marketing activities, with a treatment group that is exposed. By analyzing the difference in conversions between these groups, the model calculates the relative increase (known as lift) attributable to the marketing efforts.
To further illustrate how incremental attribution works, we can draw a parallel to medical research.
- Take 100 people – 50 get nothing or placebo, the other 50 get a new weight loss drug.
- After 8 weeks, measure how much weight has each person/group lost.
- Control (placebo) = 3.5% body weight lost.
- Treatment = 12.1% body weight lost.
- Conclusion: The treatment group experienced an 8.6% greater reduction in body weight compared to the control group.
The same logic applies to incremental marketing attribution:
- Take 100 people – 50 get nothing, the other 50 get a marketing email.
- After 3 days measure the conversion rate of both groups.
- Control = 5.2% CR
- Treatment = 7.6% CR
- Conclusion: The treatment group experienced an 2.4% greater increase in conversion rate.
This impact is what is attributed to our marketing effort, as it would not have happened without it.
Direct vs. Incremental Attribution
One of the main differences is that direct attribution can be applied anywhere, while incremental attribution is suitable only for certain uses. Scenarios where incremental attribution can be used include:
- Customer engagement (email push notifications, in-app messages, SMS)
- Website and App A/B Testing
- Conversion Rate Optimization (CRO)
These channels are all owned channels where first-party user data is available, allowing us to identify customers across multiple sessions and segment them into cohorts. This is a fundamental requirement for incremental attribution as we need a control group as a comparison baseline.
For channels like social media or paid media acquisition (Google Ads, Meta Ads), incremental attribution is simply not possible. Audiences for those channels live on third-party platforms and we don’t have an ability to assign a part of the audience as control. For example, you can’t post an Instagram story only to a part of the audience, then monitor differences in performance for those who got the post vs. those who didn’t.
Incremental attribution is not a silver bullet and comes with some downsides as well. It lacks the ability to provide multi-touch attribution information. For example, if we have a sequence of three emails, it will show the total impact, but not who each individual email contributed. Also, as incrementality is rooted in statistics, knowledge of statistics and attribution windows is a must.
Decline in Direct Attribution Efficacy
The efficacy of direct attribution has also been damaged by the rise of ad-blocking technology, privacy browsers, VPNs, and Apple’s ad tracking privacy measures introduced with iOS 14.5. All of these have some sort of negative effect on tracking and cause discrepancies in data and lower effectiveness of advertising.
How Direct Attribution can be Cheated
Direct attribution can be easily manipulated for some channels. For example, with email, the more campaigns you send and the larger your audience is, the more conversions will get attributed to your emails. This is because conversions happen all the time, so every email you send “slices” through a point in time and starts attributing conversions to itself.
Such practices might show good numbers, but they are likely not painting the right picture. Those emails might also be damaging to your brand in the long run due to over-messaging engaged users and annoying them, or messaging inactive users who cause poor email deliverablity.
The Conclusion
In the end, no single approach is superior to the others. Each method has its own benefits and drawbacks. However, in a world where direct attribution is dominant, there is an opportunity to explore incremental models. This is particularly relevant as direct attribution becomes more and more challenging due to frequent privacy changes.