¶ Sandbox PAI: Universal Mode- PMax
¶ Overview
This article explains how to use Universal Optimization with Sandbox PAI to make your Google Pmax campaigns more effective. Pmax campaigns are different from other types of campaigns—instead of Ad groups and Ads, they have Asset groups and Assets.
With Universal Optimization, you get AI recommendations for your campaigns based on the results and KPIs you want to achieve. By analyzing your campaign's past performance and configurations, Pixis AI suggests changes to your bids, budgets, and assets to help you reach your campaign goals more efficiently.
Performance Max campaigns support various assets such as Headlines, Long Headlines, and Descriptions, providing three different levers (Bids, Budgets, and Assets) to optimize these campaigns.
¶ Before we begin
-
Ensure that your Google account has AI groups with at least one Pmax campaign tagged in it.
-
This model can generate recommendations for campaigns using only the following bid strategies.
-
Maximize Conversion (Target ROAS)
-
Maximize Conversion (Target CPA)
-
Maximize Conversion Value (Target ROAS)
-
Maximize Conversion Value (Target CPA)
-
Target CPA
-
Target ROAS
-
-
This model generates asset recommendations limited to Headlines, Description and Long Headlines.
-
It only generates recommendations based on historical data from the Google Ads Dashboard. If your source of truth is a 3rd Party Platform, you are advised to use Optimize.
Note: This model won't generate bid recommendations if the Bid Strategy type is Manual CPC or Enhanced CPC.
¶ Navigate to SandBox Universal Optimization
To optimise our Pmax campaigns on Sandbox, you can follow the below steps:
-
Select PAI on the left panel
-
Select mode as Universal Mode
-
Select Universal Optimisation (UAC and Pmax) solution in Universal mode.
Note: Currently the Universal Model doesn’t have capability to generate recommendations for UAC campaigns, this feature will be added soon.
Under Universal Optimization, you have the following tabs:
-
Configuration Panel: Here, you can configure Pmax campaign details like account ID, targets, bidding strategy, etc.
-
Execution Panel Bid: You can execute the bid recommendations
-
Execution Panel Budget: You can execute the budget recommendations
-
Execution Panel Asset: You can execute the asset recommendations
-
Result Panel: You can view and download the results
-
Schedule Panel: You can schedule the recommendations
¶ Steps to Optimize Pmax Campaigns
These panels have different functions to help you Generate and Execute the recommendations. You will learn about the function of these panels in 3 parts.
-
Configuration Panel: Configuring the Optimiser to generate recommendations
-
Result Panel: Fetching and Downloading the results
-
Execution Panel Asset: Executing the recommended actions
¶ Step 1: Configuring the Optimizer to generate recommendations.
Configuration Panel: The first step in generating recommendations is to configure the optimizer. Using this panel, you can input Brand Details, Benchmarks, and Targets so that the model can generate Bid, Budget, and Asset recommendations as output.
Note: Currently, the Universal optimisation model do not support Facebook Advantage+ campaigns. This feature will be added soon.
To configure Optimizer, follow the below steps:
-
Enter Brand Name
-
Enter Brand Description
To generate product-specific or feature-specific asset recommendations, add the specific details along with the brand description prompt.
For instance: We have added the Brand Name “Real Truck” and the brand details as a description.
-
Under “Add your Google UAC/Pmax campaigns” Enter your Account ID (without Hyhens)and select Pmax.
-
Enter the conversion rate of account currency to USD to receive recommendations in account currency. If the account currency is USD, keep the conversion rate as 1
-
Under Concept ID you can select 1 or multiple campaign IDs. In case you have selected 1 campaign ID, the model will not generate budget redistribution recommendations. If you have multiple campaigns tagged under one AI group it is advised to select all the campaigns under that AI group.
¶ Build Optimisation Funnel
Under “Build Optimization Funnel” Select Bid Strategy type as per your campaigns.
¶ Select Conversion Metric
The conversion metric depends on the type of Bid Strategy you select. The conversion metrics are as followed:
-
Conversion
-
Conversion Value (Only for tROAS)
-
Conversions by Date
¶ Define Success Criteria
You can set your target goals for what you want to achieve based on specific performance indicators (KPIs). This includes setting a target return on ad spend (tROAS) and configuring your budget.
You can also input how much money you want to spend each day for the AI group. If you've chosen all the campaigns within the AI Group, you can input the total budget, and recommendations will be generated with the budget distributed among those campaigns.
Additionally, if you want recommendations to increase or decrease your budget, you can enter the minimum and maximum amounts you're willing to spend per day.
¶ Configure Optimization Levers
This section allows you to set up the levers for optimization, including Bids/Budget change thresholds and the optimization approach.
-
Enter Minimum % Bid Change Threshold: Set the smallest amount you're willing to adjust your bids or budgets by.
-
Enter Maximum % Bid Change Threshold: Decide the biggest change you'll allow for adjusting bids.
-
Select Optimization Approach:
-
Prioritize Scale Cost: Focus on growing while controlling costs.
Prioritize Scale: Concentrate on growing your campaigns.
Prioritize Cost: Concentrate on keeping costs down.
-
-
Enter Minimum % Budget Change: Set the smallest allowed change in your budget percentage.
-
Enter Miximum % Budget Change: Specify the largest allowed change in your budget percentage.
Next, click on 'Get Recommendations' and allow a few minutes for the model to process.
¶ Schedule for later
You can also schedule recommendations to be generated automatically at regular intervals. This means you won't have to keep configuring the optimizer every time.
Once the recommendations are generated you can fetch the results from Result Panel.
¶ Step 2: Fetching and Downloading the Results
Once the model has processed recommendations as per your configuration settings, you can find the results under this panel. The model processes recommendation output in 4 different .CSV files.
-
Asset Recommendations
-
Bid Recommendations
-
Budget Recommendations
-
Bid and Budget Reconciliation
You can download these .CSV files by inputting your Account ID. You will find a list of all the successfully processed outputs. Either you can click on “Download Latest” or Select a particular output and click on “Download Selected”.
Here are the sample output files for your reference.
¶ Understand Asset Recommendations
Now, let's look at the recommendations you've received. This will help you use them effectively. We'll start by understanding and using Asset Recommendations.
When you open the Asset Recommendation .CSV file, you'll see these column headers.
You can review the recommendations to decide how to improve the performance of your assets and the overall Asset AI Group.
For example: If you check these recommendations, you may find that some Descriptions in the "Hard Folding" Asset Group are labeled as "Bad" in terms of performance. In this case, the model suggests replacing these Descriptions with recommended "Variations".
Similarly, you can identify poorly performing assets in other asset groups and see the suggested replacement variations. You won't see suggested variations for assets labeled as "Good".
After reviewing all the recommendations, you can decide which assets to replace with the newly suggested variations.
We can now execute the recommendations.
¶ Step 3: Executing the Recommended Actions
The execution of asset recommendation is very a two-step process.
-
Filter the recommendations
-
Upload the file
¶ Review the recommendation & make adjustments
-
Review Before Implementation: Before making any changes to your assets, review the asset sheet to ensure they align with your campaign goals.
-
Identifying Poor Performers: In the sheet, assets marked as "bad" under the "Class" column will have replacement suggestions in the "Variation" column.
-
Option to Retain: You can choose to keep underperforming assets by removing them from the sheet.
-
Remove assets variations: As only limited no of assets are allowed for each category, you can choose which recommendations you want to add and remove the one you don’t want.
There are limits on the number of assets allowed in headlines, long headlines, and descriptions: 3-15 for headlines, 1-5 for long headlines, and 1-5 for descriptions.
Note: If you do not remove the non-selected assets the system will add the first two assets available in the cell as replacements.
¶ Upload the file
After you have made the above two changes to the sheet for all the Asset Groups. You can save the changes and upload this file using Execution Panel Asset.
-
You can drag and drop this file after selecting Publish & Execute option.
-
Next, click Submit.
With this, the assets deleted from the “Text” column will be removed from the Asset Groups and the selected recommendations available in the “Variation” column will be added as replacements.
Note: Currently Publish For Internal Review feature is inactive.
