Custom attribution models

When you use the Attribution Modeling Tool (AMT), you can use the default attribution models that come built in. However, if you want more precise control over how you distribute credit for conversions, you can also create custom attribution models. Once you create them, they're saved for reuse whenever you need them.

Set up a new custom attribution model

To use the Attribution Modeling Tool:

1. In Reporting, click the Attribution tab.

2. Click Attribution Modeling Tool in the left-hand navigation.

3. Choose a Floodlight configuration.

4. Click the first available attribution model, scroll to the bottom of the list of models, and click Create new custom model. If custom models have already been created, you can also edit them or copy them to create new custom models.

5. Name your custom model.

6. Choose a baseline model for your custom attribution model. The customization options will depend on which baseline model you choose. For example, if you choose Time Decay, you'll be able to customize the half-life.

7. In each area that you want to customize, click Off to turn on the customization options.

8. Click Save and Apply. The custom attribution model is applied to your current setup. It's also saved to the Floodlight configuration, so you can reuse it whenever you run the Attribution Modeling Tool for that configuration.

Available customizations

Depending on which baseline model you choose, you will have access to some or all of the following customizations.

• Lookback Window: Adjust how far back the attribution model will look to give credit to interactions with your ads. This option can only narrow (not expand) the existing floodlight activity lookback window for this attribution model. Otherwise, the floodlight activity lookback window will take priority. 

• Adjust credit by interaction type: Set a multiple for how much credit to give impressions versus other interactions in the conversion path. You can set a decimal value of less than 1 if you want to give impressions less credit than other interations. You can also set a multiple for impressions that precede a click by a certain amount of time. For example, you might want to give added credit to impressions that drive clicks, setting a higher multiple for impressions that come within two minutes of a click.
  
  For beta users of Twitter measurement, set custom weightings for high value social engagements and low value social engagements here. High-value social engagements extend reach to other users. Examples include shares, and retweets. Low value social engagements show interaction, but don't extend reach to other users. Examples include expands, and profile views.

• This option is not available for baseline models First Interaction and Last Interaction because these models give all of the credit to a single interaction.

• Apply custom credit rules: Set up your own custom credit rules, using a variety of specific criteria, and choose how much credit to give interactions that match the rules. For example, you can give added credit to interactions that exactly match a paid search keyword. Custom credit rules are very flexible, giving you control over how you want to distribute credit.

  If you create multiple custom credit rules that apply to the same touch point, the credit weighting for the overlapping rules is multiplied. For example, if you set rules to give a custom credit of 0.2 to a given creative and 3 to a given channel grouping, an interaction that matches both criteria would get a credit of 0.2 × 3, or 0.6.

  For baseline models First Interaction and Last Interaction, you can only use one custom credit rule because all credit is given to a single interaction.

• Set half-life of decay: If your baseline model is Time Decay, you can adjust the number of days for the half-life. An interaction that happens the number of days you set before a conversion will get half the credit of an interaction that happens the same day as the conversion.

• Specify the amount of conversion credit based on the position: If your baseline model is Position Based, you can set what percentage of credit goes to the first interaction and the last interaction, as well as what gets distributed to the middle interactions. The total must be 100%.

Example custom attribution model

In this example, you start with the Linear baseline model, which divides conversion credit evenly across touch points. A conversion path with four interactions would give each touch point 25% credit.

However, you set up a custom credit rule: for the basic channel grouping "Search," you increase the credit to 2 times other interactions in the conversion path.

When you compare the custom attribution model with the linear model for a conversion path with four touch points across four channels, you see the following results:

Example custom credit rules

Custom credit rules give you the most flexibility in applying attribution credit. Include conditions in a custom credit rule to make that rule apply to them. Exclude conditions to make that rule not apply to them. Here are a couple of examples of how and why you might use custom credit rules. 

Reduce credit for standard display

You might not want to give much credit to standard display touch points late in the conversion funnel. You figure that by that point, you've already won the customer through other marketing efforts.

To put this assumption into your custom attribution model, you set up the following custom credit rule:

Include Position in Path Exactly matching last
and
Include Basic Channel Grouping exactly matching Standard Display
credit 0.5 times other interactions in the conversion path

Reduce credit for brand keywords

You make the assumption that touch points associated with generic keywords are more valuable than those associated with brand keywords, since generic searches are from users who still need to be won over to your brand. You can reflect this assumption in your custom credit rules by setting up the following rule:

Include Paid Search Keyword Containing [brand terms]
credit 0.5 times other interactions in the conversion path

Note that you can either use a regular expression to include multiple brand terms, or you can set up multiple 'OR' statements for multiple brand terms.

Give no credit

Just like in the first example,

Include Position in Path Exactly matching last
and
Include Basic Channel Grouping exactly matching Standard Display
but credit 0 times other interactions in the conversion path.