This article provides guidance on accessing performance analysis, reviewing AI recommendations, and executing manual actions.
Once you have configured Optimized and set up automated actions, you'll start seeing its automated actions and recommendations under View Recommendations. This section provides insights into the AI's decision-making process, helping you understand its performance criteria. It enables you to evaluate whether the AI recommendations align with your campaign strategy. However, if there are instances where you prefer to override the AI's suggestions and make manual adjustments to your account, you have the flexibility to do so using manual actions. Let's explore how this process works.
You should have the Optimize feature set up. To learn more, see How to Setup Optimize?
Here, you can see the line graph on which you can map multiple KPIs to track their performance over time.
Setting up Pixis AI involves three distinct phases:
Observation: During this initial phase, Pixis AI observes the data to ensure that all data pipelines are correctly set up. It focuses on validating the integrity of the data infrastructure.
Recommendation: In the recommendation phase, users can review the AI's suggestions to determine if they align with their campaign strategy. Users can configure Pixis features to suit their preferences and ensure successful implementation of its capabilities.
Calibration: The calibration phase marks the transition to active AI involvement. Here, Pixis AI begins taking automated actions on the user's account. These actions primarily involve bid, budget, and creative rotation adjustments as configured within the optimization features.
The Action Recommendations table displays the AI Actions & generated suggestions to enhance the performance of your ad sets and campaigns. These recommendations are based on the Optimize feature under Performance AI; these actions aim to optimize key metrics, such as conversion rates and click-through rates, etc., that you’ve configured while setting up Optimize.
Additionally, you can adjust the lookback period using the calendar option and customize the displayed columns as needed. To export the data into a spreadsheet, click the download icon. For more information on manual actions, see How to Make Manual Actions.
The Action Recommendation table shows the below information:
Recommendation Date and Time: Indicates when the action was suggested by Pixis AI.
Item to Change: Specifies the Ad Set/Ad Group/Campaign targeted by the recommendation.
Recommended Action: Provides details of the recommended action proposed by Pixis.
Action Basis: Offers a brief explanation justifying the recommended action.
Action Execution Time: Displays the date and time when the action was executed on the Ads platform.
Phase: Identifies whether the recommendation is provided during the observation phase or directly implemented by the AI during calibration.
Executed Action: Describes the action executed on the Ads platform based on the recommendation.
Status: Indicates the current status of the recommended action:
Recommended: The action is suggested by Pixis.
Processing: The action is being processed by Pixis' systems.
Completed: The action has been successfully executed on the Ads platform.
Waiting for approval: This action is pending approval of the global admin. The Global admin can approve or reject this action. To learn more, see Outlier Action Limits
Rejected: The action was rejected by the Global Admin.
The Recommended Action and Action Basis are the two important columns that you should understand. They explain the AI suggestion and the basis for the action. Therefore, let’s understand these columns using an example.
Recommended Action |
Action Basis |
---|---|
Set Campaign Budget to 980.94 |
To Maintain AI Group Level Target Daily Budget Initial Budget Scaled Up USD 600.0 -> USD 1226.17 CPR Score: LOW; Result Score: LOW Performance Index: 0.0 (Rank 3/3) Final Budget: USD 1226.17 -> USD 980.94 |
Since the user manually increased the budget for the AI group by 2x, the Model scaled the budget of each asset by 2x (USD 600 -> USD 1226.17).
Then for redistributing the budget, it assessed the performance of the asset using CPR score and Result score. LOW signifies Bad performance, while HIGH is good.
It calculated a final metric called performance Index in which the asset ranks in 3rd position in the AI group (3)
The above rank signifies that the asset is performing poorly.
Based on the performance, it decreased the budget from USD 1226.17 -> USD 980.94
To learn more about Action Basis, you can see PAI Logic, which explains the action basis for each bid, budget, and creative rotation option.
In case you need to manually update the campaign budget, bid, or start/pause the campaign, click on 'Manual Action' and follow the steps below.
Select Level: Select the Level at which you want to take Action.
Campaign
Ad Group
Keywords
Ad
Select Campaign/Ad Group/Keywords/Ad: Choose for which asset you wan to take Manual Action. You can choose multiple assets within a single level. However, only a single manual action will be executed, such as setting the same budget for multiple campaigns.
Select Action: You can only perform a single Manual Action from the below option.
Bid
Budget
Change Status
Action Basis: You can add the reason behind taking Actions. These will be reflected in the Action Recommendation section on the Dashboard. Follow the same format for writing the Action Basis that the system uses.
Confidence Score: You can add the Confidence Score for the Actions.
To perform manual actions by uploading a CSV file, you can see Manual Action.