Custom bidding overview

Custom bidding lets you use your business insights and Google’s AI technology to automate how you achieve your campaign goals while maximizing your return on ad spend. 

You can reach your campaign goals by optimizing your bids for specific conversion events, goals, or impression signals that are more valuable to you. You can use custom bidding with signals from:

  • Floodlight events and custom Floodlight variables
  • Google Analytics 4
  • Impression signals
Note: Custom bidding isn’t available on YouTube and programmatic guaranteed inventory

Topics in this article

How custom bidding works

While there are standard automated bidding strategies to help you maximize campaign performance, you can use custom bidding if you need advanced control over your bidding strategy that goes beyond what automatic bidding can do.

Custom bidding lets you create a bidding strategy that automatically bids based on which impressions are most important to you.

There are 2 ways you can get started creating your custom bidding algorithm. 

  • Use rules to create custom bidding algorithms: Define custom bidding goals to optimize impression values using simple weighted conversions without having to write a script. Rules let you assign value to impressions to build a custom bidding algorithm for your campaign goals.
  • Write a custom bidding script: Use basic knowledge of python to create custom bidding scripts that let you use first-party data to optimize impression values using non-conversion based goals, such as Brand Lift measurements.
Note: You can also use the Display & Video 360 API to upload, verify individual custom bidding scripts, and assign custom bidding algorithms to campaigns.

You have granular control over your custom bidding algorithm:

  • You can use weighted values to assign higher values to impressions that are more likely to help you achieve your campaign goals and lower values to less relevant impressions.
  • You can use Floodlight tags, custom variables, Google Analytics events, or impression signals to optimize toward specific conversion events. 

Your custom bidding algorithm goes through a training phase and uses machine learning to learn from your past campaigns so it can better score and prioritize your bids on the right impressions for your campaign. 

Examples

Custom bidding algorithms can be simple or they can be complex. Here are a few examples of what you can optimize toward:

  • Brand key performance Indicators (KPIs): Such as viewability and video completions.
  • Conversion activities: Prioritize toward specific conversion activities by assigning more value to activities that are more relevant to your campaign goals.
  • Custom Floodlight variables: Such as loyalty, products, and basket size.
  • Floodlight Sales revenue: By tracking the revenue parameter using the Floodlight Sales tag.
  • Weighted conversions: Use Floodlight tracking activities that carry specific values to you depending on the product page visited.
  • Google Analytics 4: You can link Google Analytics to Display & Video 360 to define weights to events on an advertiser’s site.

Example: Brand optimization with custom bidding

When optimizing for a brand exposure, you can set your custom bidding strategy to automatically bid toward impressions that are most valuable to you using weighted values with impression data. 

For example:

You can prioritize bidding on video impressions that are more likely to increase brand exposure by creating a custom bidding goal or script that assigns:

  •  A higher value to impressions where a video was watched with the audio on
  •  A lower value to impressions where the video was watched for less than 3 seconds. 
  • A higher value for video impressions watched on connected TV (CTV).
This helps creates a custom bidding strategy that scores the priority of your video impression bids. To explore more examples, go to Using impression level data.

Example: Maximizing performance for conversion activity

When maximizing performance toward specific conversion activities, you can set your bidding strategy to automatically bid toward impressions most valuable to you using weighted values and Floodlight tags.

For example: 

You can prioritize bidding toward the most valuable Floodlight activities that are more likely to help increase sales revenue by creating a custom bidding goal or script that assigns:

  • A higher value to lower-funnel impressions, such as a completed sale.
  • A lower value to upper-funnel impressions, such as page views.

You can also use custom bidding with:

  • Floodlight Sales tags to define a revenue parameter you want to optimize toward
  • Custom Floodlight variables to optimize toward a defined basket size, product, or impressions that increase brand loyalty.
This helps create a custom bidding strategy that prioritizes activities that drive volume toward optimizing sales revenue. To explore more examples, go to Using Floodlight data.

Example: Maximizing for return on ad spend with Google Analytics and custom bidding

You can share conversion data from Google Analytics to inform your custom bidding algorithm in Display & Video 360. This lets you optimize for goals such as maximizing your return on ad spend using data from Google Analytics.

For example: 

You can optimize how custom bidding maximizes your return on ad spend by focusing on impressions that help you reach customers who are more likely to return to your website. 

You can use custom bidding and data from Google Analytics to prioritize impressions from users who click through more than 5 pages per session.

You can explore more examples by going to Using data from Google Analytics.

Permission and access

Before you begin, check that you've the permission and access you need.

For algorithms created at the advertiser level:

  • You need partner level access or the specific advertiser level access to edit the algorithm.
  • You can’t share the algorithm with other advertisers.

For algorithms created at the partner level:

  • You need partner level access to edit the algorithm.
  • You can share the algorithm with multiple partners.
  • You can view the algorithm with advertiser level access, but can’t view which advertisers the algorithm is shared with.

Training your algorithm

Your custom bidding model requires a minimum amount of impression data to learn so that it can perform well. Here are the minimum data requirements you must have for each advertiser and individual line item:

  Custom bidding goals Custom bidding scripts
Minimum data requirements for each advertiser At least 10,000 scored impressions and a minimum of 500 positively scored impressions A minimum of 500 positively scored impressions
Minimum data requirements for each line item

At least:

  • 50 positively scored impressions
  • Individual impression values must be greater than zero, and in the range of 0.000001 and 1,000,000.
  • Google Analytics require impression data for clicks.
At least 50 positively scored impressions

About the status of your custom bidding algorithm

Before you can use your custom bidding strategy on an active campaign, you’ll need to check the status of your custom bidding model to make sure it’s ready. Editing or updating your algorithm may change the status of your custom bidding model, so you’ll want to check the state of your model occasionally.

The following describes the state your custom bidding model could be in and what it means:

Status What you need to know
Training

Your model is still learning and needs time to learn.

What you need to know:

  • Assigning an untrained model to a live campaign will cause your line item to stop spending.
You may want to give your algorithm time to learn before you test its performance. As more data becomes available, it improves the accuracy of your results.
Insufficient data

Your model doesn’t have the minimum data it needs to learn. 

What you need to know:

Custom bidding uses impression data from the last 30 days, which means:

  • You may have to wait several days before you meet the minimum data requirements
  • If you pause custom bidding for more than 30 days, you may need to wait until you've met the data requirements to train your custom bidding algorithm.
  • Your model may take 1-3 days to train after you have the minimum data requirements.

You can try the following:

  • Edit your algorithm’s scoring criteria or
  • Increase the impression volume to get sufficient data to train your model.

Ready

Your model is trained and ready to use. 

What you need to know:

  • Make sure to assign your model to an active line item to prevent suspension.
Suspended

A suspended model stops training using new data. 

What you need to know

  • If no spend is associated with a model for 21 days, it gets suspended. This is to prevent unused models from taking up resources. 
  • If you need to use a suspended model, you can reactivate it by assigning your model to an active line item or insertion order set up with future budgets or flights within 21 days.
Active

Your model is assigned to an active campaign. It’s continuously optimizing and actively bidding.

What you need to know:

  • It’s important to continuously monitor how it performs and adjust your algorithm to help improve or maintain its performance.

Was this helpful?

How can we improve it?

Need more help?

Try these next steps:

true
2024 Privacy Readiness guide

Prepare for a world without third-party cookies and unlock the AI
opportunity by adopting the right durable audience and measurement
solutions.
Start Today

Search
Clear search
Close search
Google apps
Main menu
17030333230242998316
true
Search Help Center
true
true
true
true
true
69621