Channel groupings in data-driven attribution models

Channel groupings identify the types of interactions you're interested in tracking with a data-driven attribution (DDA) model, including paid search clicks driven by keywords, product groups, and dynamic targets as well as clicks from other channels.

DDA can create automated channel groupings using some basic assumptions about how most advertisers organize their search campaigns. For example, most advertisers have both upper funnel keywords and lower funnel  keywords. Starting from this assumption, DDA can distribute credit based on whether a keyword led to an upper funnel or lower funnel interaction.

If you want to model other types of interactions, you can define custom channel groupings. For example, you might want to know how your generic, brand, competitive, and promotion keywords interact together to drive purchases and newsletter signups. To create this more complex model, you specify custom channel groupings when you set up a DDA model.

Automated channel groupings

If your advertiser includes at least one Google Ads engine account, Search Ads 360 can use the account to create a DDA model. The model can be applied to other types of engine accounts in the advertiser, but Search Ads 360 needs access to a Google Ads account to create the model.

Search Ads 360 doesn't provide a way to see the channel groupings it generates, but when you use a model with automated channel groupings in a report, you'll see conversion credit distributed to keywords and other items based on the role they played in driving the conversion.

When to use automated channel groupings

Using automated channel groupings are a great choice in any of the following scenarios:

• You're just getting started with DDA and want to see how it works before you invest time in creating a more customized model
• You frequently add or change the keywords in your campaigns, and don't have time to apply labels each time you add new keywords

Automated channel groupings are not recommended if social and engine track accounts generate a large amount of traffic or conversions in your advertiser. While DDA includes interactions from social and engine track accounts when calculating conversion credit, a DDA model that uses automated channel groupings cannot accurately report the amount of credit attributed to these items. If you want Search Ads 360 reports to show the amount of conversion credit attributed to social and engine track items, use custom channel groupings instead of using automated channel groupings.

Data requirements

To automatically generate channel groupings, you need at least one Google Ads engine account in your advertiser. As long as there's at least one Google Ads engine account, Search Ads 360 can include other types of engine accounts when generating channel groupings. Your advertiser also needs to meet the minimum data requirements of clicks and Floodlight conversions.

Custom channel groupings

If you want to model specific types of interactions, you can create custom channel groupings. For example, you might create custom channel groupings that include brand and generic keywords, competitor keywords, shopping campaigns, or even ad groups that target specific RLSA audiences.

While you can specify up to 15 custom channel groupings, it's recommended that you specify no more than five channel groupings. More than that, and the model may attribute credit to less relevant channel groupings along the path. In other words, with three or four channel groupings, the distribution could be 10%, 50%, 35%, and 5%, which captures the important channel groupings in the funnel. With more channel groupings, less credit is attributed to different interactions along the path, which might not yield a good model.

For example, if you use keywords to match searches for your competitors and you want to understand how your generic, competitor, and brand keywords interact to drive conversions, create a DDA model that includes a "Competitor" channel grouping.

While it's possible to create a model with zero channel groupings, such a model behaves about the same as the basic linear model.

Why create a model with only one channel grouping?

If your main focus is understanding how much a specific group of keywords or other biddable items contributes to conversions, you can create a model with only one channel grouping. For example, if you're only interested in learning about long-tail keywords, you can create a model that only contains long-tail keywords. The model will analyze conversion paths with clicks on all types of keywords (as well as natural search clicks and display clicks if you track that data), but the model only reports incremental credit for long-tail keywords. Other keywords that were in the conversion paths will not be incremented by the model (they may be incremented by a last-click model or other model, depending on their location in each path).

For example, in this conversion path:

1. Click on "hiking boots" keyword in your "Long tail" channel grouping
2. Click on "My Brand boots" keyword (which isn't in a channel grouping)
3. Click on "My Brand X47 style" keyword (which isn't in a channel grouping)
4. Conversion

When you report on the model, you'll see some amount of credit attributed the "hiking boots" keyword. The model does not report any credit for the "My Brand boots" and "My Brand X47 style" keywords (their credit is distributed to the "Unknown" channel grouping).

If you aren't tracking natural search or display activity, it wouldn't make sense to put all of your keywords into a single channel grouping. However, if you did, the DDA model would behave like a standard linear model: it would give equal credit for the conversion to each paid search click in the conversion path.

Labels and custom channel groupings

Labels identify the keywords and other biddable items you want in a channel grouping. Since labels in Search Ads 360 are inherited, if you apply a label to a campaign and then use the label in a channel grouping, all items in the campaign will be in the channel grouping.

Similarly, if you apply a label to an ad group (instead of to a campaign), all of the ad group's biddable items will be in the channel grouping. You can also apply labels directly to keywords instead of using inheritance.

If you're creating a DDA model with custom channel groupings, apply labels to your campaigns, keywords, or other items at least 12 hours before you create the model. Otherwise, the model may not recognize your custom channel groupings. 

Multiple labels

If you apply multiple labels to a biddable item and the item ends up in multiple channel groupings, then the item's conversion credit will appear in all of the associated channel groupings. For example:

1. You apply the "Brand" label to a campaign.
2. You apply the "Brand + Product" label directly to a keyword in that campaign.
3. You create a DDA model with a "Brand" channel grouping and a separate "Brand + Product" channel grouping.
4. In a label report, the keyword's conversion count will appear in both the "Brand" label and the "Brand + Product" label. 

Changes to labels

Once you create a DDA model, you can't add labels to or remove labels from the model.

Instead of adding labels to a DDA model, you can apply labels currently used by the model to additional campaigns, ad groups, or keywords. And instead of removing labels from the model, you can remove the label from campaigns and other items. If you remove all items from a label, the DDA model will stop attributing any credit to the associated channel grouping. 

Note that a DDA model updates itself each week. Any campaigns, keywords, or other items you add or remove to your channel-grouping labels will be reviewed and incorporated into the model.

Custom channel groupings and shopping campaigns

If you want your custom channel groupings to include shopping campaigns, apply channel-grouping labels at the ad group or campaign level (labels can't be applied directly to product groups). Consider creating a separate channel grouping specifically for your shopping campaigns.

See step-by-step instructions for setting up DDA, as well as best practices for creating and using labels in a DDA model. This video is the first in a series of videos that will help you get the most out of DDA in Search Ads 360.Cross Channel Data Driven Attribution models in DS

Example custom channel groupings

Here are some examples of channel groupings that may be useful for all types of businesses:

Generic	Generic	Generic	Upper funnel	Research
	Brand only	Competitors	Mid funnel	Consideration
Brand	Brand + Product	Brand	Lower funnel	Purchase

 

Example channel groupings for verticals

 These channel groupings may be useful for specific verticals (industries):

Travel industry	Insurance industry	Retail industry	Auto industry
Generic destinations	Generic home	Generic brands	Generic luxury
Generic duration	Generic car	Generic products	Generic family
Generic types	Brand	Competitors	Generic sports
Brand	Brand + Product	Brand + Product	Brand

 

Do conversion paths need to cross through all channel groupings?

No. The model counts all combinations of conversion paths, as long as they pass through at least one channel grouping. In the example below, one conversion path passes through all three channel groupings while the other path only passes through two. In this case, you may see that each conversion attributes different amounts of credit to the channel groupings, but both paths will be counted by the model. 

Interactions outside a model's channel groupings

Even if some paid search clicks come from keywords that aren't in a model's channel grouping, when you apply the DDA attribution model to a Floodlight column and add the column to a report such as the Keywords report, you'll see the credit DDA has assigned to the keywords.

Non paid-search interactions

If a conversion path includes interactions other than paid search clicks, such as display clicks, the model assigns credit for those interactions but the credit doesn't appear in Search Ads 360 reports. For example, if a conversion path includes a display click as well as a paid search click, DDA distributes credit to both interactions, but because Search Ads 360 only reports on paid search clicks, you won't see the credit attributed to the display click in a Search Ads 360 report. 

Example of interactions outside a model's channel groupings

Consider a DDA model with two channel groupings:

• DDA Non-brand, which contains the "red shoes" keyword (as well as other keywords)
• DDA Brand + Product, which contains the "SpeedyShoes hightops" keyword (as well as other keywords)

Here's a conversion path:

1. The "red shoes" keyword matches a query and a customer clicks the related ad.
2. The "red hightops" keyword matches a query and a customer clicks the related ad.
   This keyword isn't in a channel grouping.
3. The customer clicks a display ad.
4. The "SpeedyShoes hightops" keyword matches a query and a customer clicks the related ad.
5. The customer makes a purchase, and $100 is reported as the value of the conversion.

To report on these conversions, you create a custom Floodlight column that reports revenue and apply the DDA model.

When you view conversions in a keywords report and add the custom Floodlight column (named DDA model in this example), you'll see something like this:

Keyword	Labels	Revenue	DDA model
red shoes	DDA Non-brand	0	30
red hightops	--	0	10
SpeedyShoes hightop	DDA Non-brand	100	50

Note the following:

• The standard Revenue column uses last click attribution, so all $100 of credit goes to the "SpeedyShoes hightops" keyword.
• The DDA model distributes credit for the $100 as follows:
  • $30 to the first click ("red shoes").
    Note that this is just an example, and the actual amount DDA distributes depends on what the model learns from observing your specific campaigns and conversion paths.
  • $10 to the "red hightops" keyword
  • $10 to the display ad. However, because Search Ads 360 doesn't report on display clicks, this $10 is missing from the Search Ads 360 report.
  • $50 to the last click (SpeedyShoes hightops).

Tip: Use DDA to understand the value of RLSA targeting

If you use RLSA to target high-traffic ad groups to specific audiences, add a label to those ad groups and include the label as a channel grouping in a DDA model.

The model will help you see how often RLSA targeting plays a role higher up the conversion funnel. For example, you may see that an ad group targeting a "holiday promotion" audience receives a higher amount of conversion credit from your DDA model than from a last click model. That's a good sign that this audience is highly effective at getting customers started on a conversion path. In this case, you might see more conversions if you devote more of your spend on this audience. 

Note that you'll need an ad group with a high number of clicks. If the traffic is low, the DDA model won't have enough data to point out any trends.