The goal of a marketing attribution platform is to enhance the ability of marketers to optimize the channels and campaigns that are working, adjust the ones that aren’t, and highlight where there are opportunities to increase the return on marketing investments.
Attribution platforms do this by tracking and crediting the interactions on a website that lead to conversions.
Conversions can be any event the user defines as important to the company’s objectives, e.g., filling in a form, booking a demo, or completing a purchase. In the e-commerce world the most important conversion event is of course a purchase by a customer, as revenue is the key measure of success for a store.
In the table below we summarize the pros and cons of the more widely used attribution models above.
Model | What User Interaction (Touchpoint) Is Measured | Pros | Cons | Dimensions Attribution Enhancement to Standard Model | |
---|---|---|---|---|---|
Dimensions Machine Learning | Developed by AdAmplify using its Machine Learning (ML) algorithms, the weighting of channels or campaigns is automatically calculated based on the behavior of customers as they make purchases on a site. | As the weight of each channel or campaign is dynamically calculated each day based on the actual behavior of users leading to purchases, the ML model accurately reflects the real contribution of each channel in contributing to purchases. | None | Based on AdAmplify’s proprietary Machine Learning algorithm, Dimensions calculates the probabilty of purchases for each channel. As user behavior is different on each store’s website, this model is in essence a “custom” model generated for each store. | |
Linear | Equal weight or credit is assigned to each touchpoint leading to a conversion. This model relies on a general “rule of thumb” assuming that all touchpoints are equal and is thus a “heuristic” model (a “guesstimate”). | Gives credit to all touchpoints equally. | All touchpoints are not equal, some play a more important role in conversions. This model also assumes all sites have the same user behavior, which is not the case. | Purchases are attributed based on the touchpoints leading to initial or subsequent purchases. | |
Linear Non-Direct | Equal weight is assigned to each touchpoint leading to a conversion. Direct touchpoints are disregarded unless Direct is the only touchpoint. (Direct touchpoints are those from a bookmarked site or when a user types the site’s URL into a browser’s address bar). | By disregarding most Direct touchpoints, focuses on the touchpoints that a merchant can influence in their marketing. | All touchpoints are not equal, some play a more important role. | Because Direct touchpoints cannot be influenced by your marketing efforts, this model shows the power of the channels or campaigns that can actually be influenced/optimized by your initiatives to attract visitors who over time become loyal customers. | |
Custom Attribution Models | The user assigns a weight to each touchpoint which determines how the different touchpoints are assigned credit for conversions. | Offers the user some ability to adjust the weight of different touchpoints for potentially something more “realistic”. | Difficult & time-consuming to manually determine how much credit to assign to each touchpoint. Weighting is dynamic, changing over time so this would require constant manual re-calculation. | Not supported by Dimensions. Our Machine Learning Attribution Model automatically and dynamically recalculates the weight for each site based on that site’s customer behaviour (in essence, automatically building a “custom model” for each store). | |
First Click | Revenue is attributed to the first touchpoint (e.g., Paid Search channel). | Shows the power of a channel or campaign in attracting first time purchasers. | Disregards other touchpoints that may have contributed to the purchase (6-8 touchpoints are typical for many user journeys leading to purchase). Ignores the value of later touchpoints that contributed to a conversion. | Repeat purchases are attributed to the first click touchpoint (channel or campaign), regardless of the last click’s touchpoint. Attributing repeat purchases to First Click allows you to see the power of the channel or campaign in attracting initial purchasers who subsequently become loyal customers. | |
Last Click | Revenue is attributed to the last touchpoint (e.g., Paid Search channel or campaign). | Shows the power of a channel or campaign in contributing to the purchase conversion. | Disregards other touchpoints that may have contributed to the purchase; may overstate the value of some channels. | Each purchase, initial or subsequent, is attributed to the last click touchpoint (channel or campaign) from which the purchaser came. Attributing repeat purchases to Last Click allows you to see the power of each channel or campaign in converting your repeat purchasers. | |
Last Non-Direct Click | Revenue is attributed to the last touchpoint but Direct touchpoints are disregarded unless Direct is only touchpoint. | By disregarding most Direct touchpoints, focuses on the touchpoints that merchant can influence in their marketing. | Disregards other touchpoints that may have contributed to the purchase; may overstate the value of some channels. | Because Direct touchpoints cannot be influenced by your marketing efforts, this model shows the power of the channels or campaigns that can actually be influenced/optimized by your initiatives to attract visitors who over time become loyal customers. | |
Position Based | 40% of the conversion is credited to the first and last touchpoints; the balance is divided up evenly among the other touchpoints. | An attempt to create a more realistic model than the linear model. | Overvalues the first and last touchpoints, and undervalues the intermediate ones. | Not supported by Dimensions. As is the case with the other standard multi-touch models shown here, Dimensions Attribution’s Machine Learning model is a superior representation of actual channel weights. | |
Time Decay | The most recent touchpoint receives the most credit and the older ones progressively less. | More representative than Last Touch as it does distribute credit to earlier touchpoints. | Overvalues the last touchpoint, and undervalues less recent touchpoints (such as the first click) that may have contributed to the conversion more than this model gives credit. | Not supported by Dimensions. As is the case with the other standard multi-touch models shown here, Dimensions Attribution’s Machine Learning model is a superior representation of actual channel weights. | |
* Note: where applicable, Dimensions allows you to drill down into ad groups or ads runnning under your campaigns. |
See also: Which Attribution Model Should I Use?