Adobe Analytics: Attribution IQ Models
This is a classic case of an image paints a thousand words. This subject has come up a a lot lately due to the availability of the Attribution IQ (or what I now call “AIQ”) functionality, as part of the Adobe Analytics Summer 2018 release.
To help illustrate what AIQ is actually doing under-the-hood, I produced this simple overview within Analysis Workspace, just to help showcase what each of the models is actually doing on a single success/conversion event.
It should hopefully make sense to those new to attribution modelling. Ps. Remember it works on any dimension and success event combination!
Model Explanations
= = = =
First Touch
100% credit to first touchpoint.
When it’s useful: This model is appropriate if you run ads or campaigns to create initial awareness. For example, if your brand is not well known, you may place a premium on the keywords or channels that first exposed customers to the brand.
This is another common attribution model useful for analysing marketing channels intended to drive brand awareness or drive customer acquisition. First Touch is frequently used by Display or Social marketing teams but is also great for assessing onsite product recommendation effectiveness.
= = = =
Last Touch
100% credit to last touchpoint.
When it’s useful: If your ads and campaigns are designed to attract people at the moment of purchase, or your business is primarily transactional with a sales cycle that does not involve a consideration phase, the Last Interaction model may be appropriate. This is the most basic and common attribution model and is frequently used for conversions with a short consideration cycle. Last Touch is commonly used by teams managing search marketing or analyzing internal search keywords.
= = = =
Linear
Equal credit split across all touhpoints.
When it’s useful: This model is useful if your campaigns are designed to maintain contact and awareness with the customer throughout the entire sales cycle. In this case, each touchpoint is equally important during the consideration process.
This model is useful for conversions with longer consideration cycles or user experiences that need more frequent/consistent customer engagement. Linear attribution is often used by teams measuring mobile app notification effectiveness or with subscription-based products.
= = = =
Participation
All touchpoints receive 100% credit.
When it’s useful: When you want to understand which campaigns are actually part of the customer journey. Other models may allocate zero credit, whereas this model at least recognises every touchpoint in full. Useful for overlap analysis. This model is excellent for analysis and discovery to understand how often your end users or customers are exposed to any particular channel, page, or interaction. Media teams will often use this model to calculate content velocity, and retail organisations will often use this model to understand which parts of their app or website are on the critical path of conversion.
= = = =
Same Touch
100% credit to the hit/event/measurement that contained the success event.
When it’s useful: This is a helpful model when evaluating the content or user experience that was presented immediately at the time of conversion. Product or design teams will often use this to assess the effectiveness of a page where conversion occurs. A simple example would be to test which day of the week a conversion took place on, versus the participation of all days of the week interactions that led up to the conversion; highlighting which day users visit in the lead up to a conversion.
= = = =
U-Shaped (aka. Position Based)
40% to first, 40% to last, 20% split across all remaining touchpoints.
When it’s useful: If you most value touchpoints that introduced customers to your brand and final touchpoints that resulted in sales. This is a great model for those who value interactions that occurred first or last (introduced or closed) in a conversion, but still wish to recognise the assisting interactions. U-Shaped attribution is often used by teams who take a more balanced approach but want to give more credit to channels that found or closed a conversion.
= = = =
J-Curve
60% credit to the last interaction, 20% credit to the first interaction, and shares the remaining 20% among any interactions in between.
In attribution look-backs with only one touchpoint, 100% credit is given to the single touchpoint, and in cases with only two, 75% is given to the last touchpoint, and 25% is given to the first.
When it’s useful: Similar to U-shaped, this is a great model for those who value interactions that occurred first or last (introduced or closed) in a conversion, but which to emphasise the interaction that closed on the conversion. J-Shaped attribution is often used by teams who take a more balanced approach and want to give more credit to channels that closed a conversion
= = = =
Inverse J-Curve
60% credit to the first interaction, 20% credit to the last interaction, and shares the remaining 20% among any interactions in between.
In attribution look-backs with only one touchpoint, 100% credit is given to the single touchpoint, and in cases with only two, 75% is given to the first touchpoint, and 25% is given to the last.
When it’s useful: Similar to U-shaped, this is a great model for those who value interactions that occurred first or last (introduced or closed) in a conversion, but wish to emphasise the interaction that initiated the conversion. Inverse J attribution is often used by teams who take a more balanced approach and want to give more credit to channels that initiated a conversion.
= = = =
Time Decay
The Time Decay model follows an exponential decay with a custom half-life parameter (default is seven days).
The weight of each channel depends on the amount of time that passed between the touchpoint and the eventual conversion and is determined by the formula 2^(-t/half-life) where t is the amount of time between a touchpoint and conversion.
When it’s useful: If you run one-day or two-day promotion campaigns, you may wish to give more credit to interactions during the days of the promotion. In this case, interactions that occurred one week before have only a small value as compared to touchpoints near the conversion. Also great for promotions that have a hard stop date, such as Black Friday or Christmas!
This is a good model for teams that run promotions across a predetermined number of days who wish to emphasise channels that occurred more recently. Time Decay attribution is often used by teams running video advertising or teams scheduling their marketing around a significant event with a predetermined date (such as a conference or sporting event).
= = = =
Custom
User defined %’s for first, last and middle split of touchpoints.
When it’s useful: If your organisation is uncomfortable with the defaults provided by Adobe Analytics, a custom model allows you to specify the weights that make the most sense to your organisation.
= = = =
Further reading: https://marketing.adobe.com/resources/help/en_US/analytics/analysis-workspace/attribution.html
