¶ Sandbox TAI: Universal model (Bid, Budget, Targeting)
In this article, we will learn how to configure the Universal model to generate and execute bid, budget, and targeting recommendations for Facebook ad accounts.
The Universal Model in the GAI Solution Suite leverages AI to identify underperforming ad sets and provides recommendations for improving bid, budget, and targeting strategies. This helps ensure that your Facebook campaigns are optimized for better performance.
The following sections help you to configure which campaigns to optimize, set target results and cost metrics, generate audience cohorts, and more.
-
Which assets should Pixis AI optimize?
-
How should AI Optimize these assets?
-
How should Pixis evaluate and optimise targetings for these assets?
-
How should Pixis AI define the success criteria?
-
How should Pixis AI use different levers to optimise?
¶ Which Assets should Pixis AI Optimise?
In the Select the Platform field, we only provide support for Facebook Ads.
Account ID:
-
Enter your Facebook Ads account ID.
-
This ID is necessary to pull the correct data for analysis.
Select the Optimization Level
This field will determine the ad sets on which the recommendations will be generated:
-
Sub Optimizer ID
-
Concept ID
¶ Sub-Optimizer ID
This is the element ID for your AI group, which can be found on the admin panel at the AI group level. Choose the Sub-Optimizer ID/Element ID associated with your ad account. Recommendations will be generated for all campaigns/ad sets within this ID.
¶ Concept IDs
-
This is the ad set ID for the specific ad sets you want to review. It allows for a more targeted analysis of individual ad sets. Multi-select the ad sets and campaigns for which you want to generate recommendations.
Note: To run the model at an AI group level, select Sub-optimizer ID (your element ID). For specific ad sets, select Concept ID. Scheduling is only available at the element/sub-optimizer level.
¶ How should Pixis AI optimise these assets?
¶ Select Result Metric and Cost Metric
-
Select Result Metric:
-
Choose the metric that represents the desired outcome of your ads. For example, Select "Clicks" if you want to optimize for the number of clicks your ads receive.
-
-
Select Cost Metric:
-
Choose the metric that represents the cost associated with achieving the result. For example, select "Cost per Result" to optimize the cost per click or conversion.
-
Note: Currently, you can only select default Result & Cost Metrics. In future updates, you will be able to select your custom Result and Cost metic.
¶ Add Result Metric Details for Sub-optimizer
-
Select Source for Result Metric:
-
Specify the source data for the chosen Result Metric. For example, select "Link Clicks" if you want to track and optimize based on clicks on your ad links.
-
-
Attribution Setting:
-
Choose the attribution setting that determines how conversions are credited to your ads. For example, select "default" to use the standard attribution settings for your account.
-
¶ How should Pixis evaluate and optimise targetings for these assets?
In this section, you can set benchmarks for evaluating and optimizing the targetings for your Facebook ad sets.
¶ Add Benchmarks to Evaluate Targeting for Sub-optimizer
-
Select Results Custom Audiences: Choose custom audiences from the dropdown that represents the desired results. These are lower funnel custom audiences, crucial for evaluating past performance and targeting accuracy.
-
Select Clicks Custom Audiences: Select custom audiences that have shown engagement by clicking on your ads. These are top funnel lookalike audiences, important for assessing potential reach and engagement.
-
For example, Select a custom audience of users who have clicked on your ads but not necessarily completed a purchase.
-
-
Target Conversion Rate (CVR): Define the desired conversion rate for flagging underperforming ad sets.
-
Target CTR (Click-Through Rate): Set the target CTR to identify ad sets that are not performing well in terms of engagement.
-
Target Cost per Result (CPR): Define the acceptable cost per result to maintain cost efficiency.
¶ How should Pixis AI define the success criteria?
In this section, you can configure different spending criteria and budgeting options for effectively managing campaign performance.
¶ Select Spending Criteria:
-
Allow Gradual Scaling Up/Down: You have the option to implement gradual adjustments to your budget rather than sudden changes to avoid disrupting campaign performance.
-
Maintain Daily Target Budget: You can set a daily target budget to ensure tightly monitored spending.
-
Maintain Current Daily Budget: Users can choose to maintain a fixed budget allocation to provide stability and predictability for their campaigns.
¶ Set daily budget and targets
-
Enter Min/Max Daily Budget: You can configure minimum and maximum bid thresholds for optimal performance.
-
Enter Daily Target Result: Users can set realistic and achievable targets based on selected metrics for campaign objectives and historical data.
-
Enter Target Cost: You can define the target cost per result to optimize cost-effectiveness for acquiring customers.
Note: While adding Target Conversions Rate and CTR, you have to divide the number by 100 and then enter in the field.
¶ Select Optimization Approach:
-
Scale First: This approach focuses on performance scaling rather than optimizing cost.
-
Cost First: This approach gives more importance to cost optimization rather than scaling performance.
-
Balance Both: You can opt for a balanced approach under a manual or AI-driven method to equally prioritize both scale and cost.
¶ How should Pixis AI use different levers to optimise?
You can set a minimum and maximum budget, as well as a bid change percentage.
Schedule Recommendations:
-
Add email addresses (comma-separated for multiple emails) to receive scheduled recommendations.
-
This ensures you get timely updates and can take action promptly.
-
Click on 'Generate Recommendations.'
To generate recommendations, the system performs an analysis that focuses on comparing your ad sets' performance against various benchmarks:
1. 7-Day Metrics Comparison: This compares the performance of your ad sets over the last 7 days with the AI group's performance over the same period. For example, if your ad set had a click-through rate (CTR) of 2% and the AI group's average was 3%, this would be highlighted.
2. 28-Day Metrics Comparison: Similar to the 7-day comparison but over a longer period. This helps identify longer-term trends and performance issues.
3. User Input Benchmarks: If you've provided specific benchmarks (like a target CTR or conversion rate), the platform will compare your ad sets against these benchmarks. For instance, if your target CTR is 4% and your ad set is only achieving 3%, this will be flagged.
After analyzing the data, the system will provide a download recommendation button. You can then download the Google Sheet.
¶ Execution Panel
You can execute these recommendations for your Ad account in the Execution Panel by uploading the recommendations sheet. You can execute the following actions:
-
Targeting Recommendations: These are suggestions on how to better target your audience. For example, if your current targeting isn't performing well, the platform might suggest new demographics or interests to target.
-
Bid-Budget Recommendations: These are suggestions on how to adjust your bids and budgets to optimize performance. For instance, if your current budget is too low to achieve your goals, the platform might recommend increasing it.
Here’s a sample recommendation sheet for the Bid+budget and targeting execution.
¶ Execute targeting recommendations from the execution panel
After downloading the output recommendation sheet, you can review the underperforming ad sets. The system flags the cohorts it analyzes as not performing well and provides options to replace them. To quickly implement the suggestions, check the underperforming ad sets under the "adset_review_summary" worksheet. Then, under the "execute_published_cohorts" worksheet, you’ll find the targeting recommendations that you can execute from the execution panel.
¶ Adset_review_summary worksheet
To know which ad sets are flagged, you can check the columns below:
-
Bad performance alarm: This indicates low-performing ad sets.
-
Targeting fault: There’s a fault in targeting and it suggests changing or updating the targeting.
Once you know which ad sets are underperforming, you can go to publish_adsets_cohorts to check the recommendations cohorts and execute them.
¶ Publish ad sets worksheet.
The platform suggests ten new targeting options to replace the faulty ones. For example, if your current targeting isn't performing well, the platform might suggest new demographics or interests to target.
Audience Relevance Score: Each new targeting option comes with an audience relevance score, indicating how well it matches your goals. For instance, if the old ad set had a relevance score of 0.4, the new options might have scores of 0.5 or higher.
Here, you can review the O, P, & Q columns to change the status of the cohorts from FALSE TO TRUE to publish them and set daily and lifetime budget.
Important notes:
- One Sheet at a Time: Ensure you upload and execute one sheet at a time. If you need to make additional changes, use a different sheet.
- Budget Consistency: If you want the new ad sets to have the same budget as the reference ad set, leave the budget columns empty. The system will automatically apply the same budget.
- Avoid Duplicates: When using the same sheet for different actions, ensure previously executed actions are set to "false" to avoid duplicates.
¶ Pause ad sets
You can decide if you wish to pause the current flag ad set or not.
¶ Once the sheet is uploaded
You will see what actions are about to be executed for each reference ad set where you made changes in Excel.
On the left side, you can view whether any pause action has been taken for the reference ad set.
On the right side, under "Cohorts to be launched with budget," you will find the new cohorts to be launched along with their respective budgets.
After reviewing, you can press submit, and a confirmation screen will appear. This screen will provide an overall summary for each reference ad set, indicating whether a pause action has been taken and detailing the cohorts that are being published for the ad set.
Note: Each cohort will be launched as an individual ad set.
Once you verify all the actions to be executed, you can Accept and Confirm.