Use rules to create custom bidding algorithms

You can use rules to optimize for weighted conversions, clicks, brand awareness, or define your own custom metrics. This improves how custom bidding prioritizes impressions that align with your campaign's objective, without having to write a script.

Set up

1. Go to Resources Custom Bidding.
2. Choose New algorithm.
3. Enter a name for your algorithm.
4. Select your objective. The objective you choose determines which rules can be selected. If you prefer to use scripts, select the objective as Custom. See Create and use a custom bidding script for more information about the Use script option. 
5. If you’re creating the algorithm at the partner level, select an advertiser.
6. Set up rules based on the objective you selected:
   • Conversion rules:
     • Click Add rule and choose a conversion type. Choose one or more Floodlight activities or Google Analytics conversions and click Select.
     • Set the following rule parameters:
       • Conversions to count: 
         • Total conversions:  Count every conversion from an impression. 
         • Unique conversions: Count only 1 conversion per impression, even if someone converts multiple times.
         • Sales quality: Calculates the impression value by the number of sales. 
         • Sales revenue: Calculates the impression value by the sales revenue.
       • Conversion type: Available when setting goals using Floodlight activities. Determines which conversions count toward the impression score. 
       • Score: Score is a numeric value that you assign to a rule. The score will apply to the impression that satisfy the rule. Values must be > 0. Up to 3 decimal points are allowed.
   • Click rules: We recommend combining click rules with multipliers. If you only care about clicks, consider the Maximize clicks automated bid strategy.
   • Brand awareness rules: Scores can be determined by looking at your Display & Video 360 reporting or using insights from other sources to assess the relative value of signals like device type, viewable, time on screen, ad position, and time of day.
7. Choose how to Score when more than one rule is true. 
8. Select a conversion attribution model:
   It's important to monitor the data you provide to attribution models. Inaccurate or unstable data may result in unexpected bidding outcomes.
   • For goals using Floodlight activities, choose one of the following:
     • Floodlight attribution model (primary): An attribution model you've previously set as the primary. A primary attribution model is the default model for all line items you associate with the Floodlight. For more information, go to change primary attribution models.
     • Linear: A default attribution model which gives equal credit to each channel a user interacts with on their way to a conversion.
     • Pay-per-active view (read-only): The default attribution model if using pay per active view or clicks.
     • Base last interaction: A default attribution model which gives 100% credit to the last channel the user interacted with.
     • Data-driven attribution (DDA) model: A model you create in Campaign Manager 360. You can train data-driven attribution models using your own consumer behavior data. You should consider this option if you're new to attribution. For more information, go to create a new data-driven attribution model in Campaign Manager 360.
   • For goals using Google Analytics events, choose:
     • Google Analytics Last Click, Non-fractional Credit: A model which ignores direct traffic and attributes 100% of the conversion value to the last channel clicked before buying or converting. About the default MCF attribution models.
9. (Optional) Set up multipliers for impressions: Multipliers are in addition to rule scores. For example, an advertiser can multiply the scores for a certain device, time of day, ad position, etc. 
   • Click Add multiplier. 
   • Choose a multiplier: The multiplier  value can be any decimal number greater than 1 and less than or equal to 5. The default multiplier is 1 when no multipliers are added. 
10. Click Save as draft to move on to testing.

Test

Testing your algorithm generates a chart that shows the distribution of scores for your chosen impressions. You can use this data to check if your custom rules perform as expected.

Under Sample type, choose how impressions are sampled: 

• All impressions: Recommended to understand overall expected distribution of impression values.
• Only impressions with clicks: Useful if clicks are central to the script.
• Only impressions with conversions: Useful if conversions are central to the script or rules you’ve added. This sample includes all conversion events in your advertiser and will be based off the default attribution model 

You can test your algorithm using a random sample of 10,000 impressions from eligible impressions. You can review the following results:

• The number of impressions in the sample: The number of sampled impressions included in the results.
• Impression value/cost: The impression value/cost for the sampled impressions.
  Note: Included in the results only when using all impressions as the sample type.
• Percentage of execution errors: The percentage of sampled impressions that returned an error after testing the script.
• Percentage impressions scored: The percentage of sampled impressions successfully scored.
• Percentage of positive impressions scored: The percentage of sampled impressions  successfully scored with a score over 0.

You can download a .csv file to view the details of individual values in the results. 

Save

After you create your script, it requires time to process. It may take 20 mins before it's available. When it's available, you can view your script from the script list.

For more information about the default attribution models, go to about attribution modeling.