By Zach Schapira on March 9, 2017
We’re excited to announce the release of a powerful new tool to help you completely customize and experiment with your rules-based attribution models.
Using the new Model Builder, an easy step-by-step interface, you can take any base attribution model (e.g. even, position-based or time decay) and modify it by changing the settings both for how the conversion path is formed, as well as how credit is distributed across different touchpoints on that conversion path (more on what each means below).
Then, once you’ve customized your new rules, you can choose to either:
- Run your attribution model as a one-time test to check results,
- Set it to recur alongside your preferred model for up to 30 days (you can even A/B test up to 10 models simultaneously), or
- Select it to entirely replace your preferred model.
Customize the conversion path
Having this flexibility is important because out of the box, rules-based models aren’t versatile enough to account for unique preferences and circumstances across brands, products, and companies.
For instance, when customizing how the conversion path is formed, the Model Builder will ask you questions such as:
How long do you want your lookback window to be?
In other words, do you want to consider an impression or click to be part of the conversion path if it happened more than a month prior to the conversion?
How long do you want the re-processing period to be?
In other words, how long after the conversion is recorded should the system continue to update the original path as new or delayed data comes in? This could be as a result of, for example, latent data feeds, corrected data connection errors, or new cross-device insights.
Do you want to trim the path based on customer inactivity?
For example, if your lookback window was three weeks but 2.5 weeks passed between your video ad and the next click on branded search, should the video ad still be part of that conversion path?
What level of cross-device identity matching do you want applied?
For example, do you want to leverage data only from your account, from the Impact Radius consortium, or from one of our third party integrations?
Customize credit attribution
Alternatively, when customizing how to attribute credit across each conversion path, the Model Builder will ask you questions such as:
Do you want to group events that happen very close to each other into a single touchpoint?
For example, if you have a conversion path (even model) where 5 impressions appeared over the span of two hours, should each of those impressions count equally or be grouped as a single event?
Are there certain touchpoints that you know deserve more credit than others?
For example, you might want to weigh a click as 10 times the value of an impression. Or you might have a revenue partner who consistently drives high AOV (average order value) because of the unique way he promotes your brand, so you want to give him double the credit of whatever the model would usually assign.
Customize base settings
You can also adjust the base model settings for position-based models and time decay models, adjusting the value assigned to a specific position or the rate at which value decays.
While an algorithmic model is the most reliable method for attribution (the algorithm learns from your data and would figure out these answers for you), you can drastically improve the effectiveness of rules-based models just by pulling the right levers in the right places.