OurAttribution Models

Last Touch Modeling



Last touch is by far the most prolific of all the attribution models. Most marketers and businesses have undoubtedly used this method before and while its shortcomings have been widely publicized, it has endured for two reasons: 1.) last touch does give you a rough look at your digital marketing channels and their performance and 2.) last touch is by far the easiest attribution model to calculate. Below is an example of a conversion that occurred during the fifth touchpoint in a customer journey.

Last Touch Attribution Model Example

If you are looking simply for the marketing channels that had an immediate, direct impact on a transaction, the last touch model works; however, this model fails to capture any upstream interaction that occurred in the customer funnel and can lead to inaccurately drawn conclusions from the data.

First Touch Modeling



The First Touch model attributes 100% of conversion credit to the first interaction a customer has with a site regardless of whether there are 2 touchpoints or 30. Below, you can see a graph depicting a customer journey with 5 touchpoints with the transaction occurring in the final touchpoint.

First Touch Attribution Model Example

This model can certainly have its drawbacks, but for particular business types the first touch model can be invaluable. In B2B sales businesses with long customer journeys and numerous offline touchpoints with sales teams, the first touch model is great at helping to discern the most valuable digital marketing channels. One caveat with most attribution systems is that the lookback windows only go back 30, 60 or 90 days. If a customer journey to a transaction is relatively short this time frame is fine, but if it exceeds the lookback window, you are no longer getting any actual first touch channel data, but rather the first touch that occurred during the lookback window. At oobyMetrics, we have removed lookback windows to remove this drawback to the first touch model and any of our other multi-channel attribution models.

Last Non-Direct Modeling



The Last Non-Direct Model, in an attempt to fix one of the underlying issues with Last Touch attribution modeling, gives 100% credit to the last non-direct touchpoint. Below is a graph depicting a customer journey with 5 touchpoints, the last touchpoint being from a direct channel.

Last Non-Direct Attribution Model Example

Although remarkably similar to the last touch model, last non-direct modeling corrects for issues that arise from the "catch-all" bucket of direct touchpoints. The term "direct" is meant to represent customers who enter your url directly into their browser; however, the direct channel ends up catching traffic that is either untagged or improperly tagged which can create data inaccuracies. The Last Non-Direct model is great if your Direct channel data is causing problems and providing unreliable information.

Linear Modeling



The linear model is a multi-channel attribution model and is the simplest of the multi-channels to calculate. Every touchpoint in the path to conversion is counted and given an equal weighting. For instance, if there were 5 touchpoints with different channels, each touchpoint/channel would be credited 20%. An example of this linear attribution weighting can be found in the chart below.

Linear Attribution Model Example

Depending on the business and acquisition strategy, this can be a useful model, but it does have some drawbacks. If each of your channels does an equal amount of the heavy lifting to create a conversion, the linear model is perfect; however, if you have channels that do more "work" than other channels, linear attribution (while more accurate than last touch) can be misleading. For example, if your acquisition strategy is heavily front loaded with an intensive lead generation channel, such as trade shows but it also uses email campaigns that lead to conversions. In this instance if it took 9 emails before a conversion, the tradeshow channel would only receive 10% of the credit, even though you may feel it did 40% of the work.

Positional Modeling



The Positional model goes by a number of names like Position-based or U-Shaped modeling. It is another multi-channel attribution model combining the benefits of first touch and last touch. In this model, every touchpoint gets some credit, with the most going to the first touchpoint (40%) and the last touchpoint (40%). The remaining 20% of credit gets split up evenly between the remaining touchpoints. Below is a chart showing a customer journey consisting of 5 touchpoints, the first getting 40% of the credit, the middle three touchpoints each getting 6.67% of the credit, and the last touchpoint getting 40% of the credit.

Positional Attribution Model Example

The positional model can work exceptionally well depending on your business needs. In a business where leads matter as much as the conversions, the positional model excels while still attributing some credit to channels that assisted along the way. To increase the accuracy of this model, oobyMetrics removed lookback windows to ensure that the heavily weighted first touch channel is truly the first channel, not the first channel within the lookback window.

Time Decay Modeling



The Time Decay attribution model is a multi-channel / multi-touch attribution model focused on giving credit to all touchpoints within the customer journey, but more heavily favoring the touchpoints closest to the conversion. The graph below shows a 5 touchpoint customer journey with a conversion at the fifth touchpoint. As you will note, each subsequent touchpoint gets slightly more weight than the one before.

Time Decay Attribution Model Example

As with all attribution models, the Time Decay model works well for certain business needs. For example, the Time Decay model is beneficial for an e-commerce website running multiple digital marketing channels because the initial touchpoint does not necessarily deserve more credit than the others. In this instance you want to most heavily weigh the touchpoint directly influencing the conversion. At oobyMetrics, we have removed the lookback window to ensure all touchpoints are included, which in turn creates more accurate data for our users.



Get Started