Customer AI error troubleshooting
Customer AI displays errors when model training, scoring, and configuration fails. In the Service instances section, a column for LAST RUN STATUS displays one of the following messages: Success, Training issue, and Failed.
In the event that Failed or Training issue is displayed, you can select the run status to open a side panel. The side panel contains your Last run status and Last run details. Last run details contains information on why the run failed. In the event that Customer AI is not able to provide details on your error, contact support with the error code thats provided.
{width="300"}
Unable to access Customer AI in Chrome incognito
Loading errors in Google Chrome’s incognito mode are present because of updates in Google Chrome’s incognito mode security settings. The issue is actively being worked on with Chrome to make experience.adobe.com a trusted domain.
{width="500"}
Recommended fix
To workaround this issue you need to add experience.adobe.com as a site that can always use cookies. Start by navigating to chrome://settings/cookies. Next, scroll down to the Customized behaviors section followed by selecting the Add button next to “sites that can always use cookies”. In the popover that appears, copy and paste [*.]experience.adobe.com
then select the Including third-party cookies on this site checkbox. Once complete, select Add and reload Customer AI in incognito.
Model quality is poor
If you receive the error “Model Quality is poor. We recommend creating a new app with the modified configuration”. Follow the recommended steps below to help troubleshoot.
{width="300"}
Recommended fix
“Model quality is poor” means that the model accuracy is not within an acceptable range. Customer AI was unable to build a reliable model and AUC (Area under the ROC curve) < 0.65 after training. To fix the error, it is recommended that you change one of the configuration parameters and rerun the training.
Start by checking the accuracy of your data. It is important that your data contains the necessary fields needed for your predictive outcome.
- Check whether your dataset has the latest dates. Customer AI always assumes that the data is up-to-date when the model is triggered.
- Check for missing data within your defined prediction and eligibility window. Your data needs to be complete with no gaps. Also make sure your dataset meets the Customer AI historical data requirements.
- Check for missing data in commerce, application, web, and search, within your schema field properties.
If your data does not seem to be the problem, try changing the eligibility population condition to restrict the model to certain profiles (for example, _experience.analytics.customDimensions.eVars.eVar142
exists in last 56 Days). This restricts the population and size of the data used in the training window.
If restricting the eligibility population did not work or is not possible, change your prediction window.
- Try changing your prediction window to 7 days and see if the error continues to occur. If the error no longer occurs, this indicates that you may not have enough data for your defined prediction window.
Errors
Suggested solutions:
1. Check data availability
2. Decrease the prediction goal timeframe
3. Modify the prediction goal definition to include more users (Error code: VALIDATION-400 NOT_ENOUGH_OBJECTIVE)
Suggested solutions:
1. Check data availability
2. Decrease the prediction goal timeframe
3. Modify the prediction goal definition to include more users. (Error code: VALIDATION-400 NOT_ENOUGH_OBJECTIVE)
Suggested solutions:
1. Check data availability
2. If an Eligible population definition is provided, decrease the eligibility filter timeframe 3. If an Eligible population definition is not provided, try adding one (Error code: VALIDATION-401 NOT_ENOUGH_POPULATION)
Suggested solutions:
1. Check data availability
2. If an Eligible population definition is provided, decrease the eligibility filter timeframe.
3. If an Eligible population definition is not provided, try adding one. (Error code: VALIDATION-401 NOT_ENOUGH_POPULATION)
Some suggestions include:
1. Modify your configuration to add an eligible population definition.
2. Use additional data sources to improve model quality
3. Add custom events to include more data in the model (Error code: VALIDATION-402 BAD_MODEL)
Some suggestions include:
1. Please consider modifying your configuration to add an eligible population definition.
2. Please consider using additional data sources to improve model quality. (Error code: VALIDATION-402 BAD_MODEL)
Some suggestions include:
1. Please make sure the model is trained with recent data, if not, consider retrain your model.
2. Please make sure there is no data issue (such as missing data/data delay) in scoring tasks. (Error code: VALIDATION-403 INELIGIBLE_SCORES)
Some suggestions include:
1. Please make sure the model is trained with recent data, if not, consider retrain your model.
2. Please make sure there is no data issue (such as missing data/data delay) in scoring tasks. (Error code: VALIDATION-403 INELIGIBLE_SCORES)
We require 120 days of recent data. For more information, please check the data requirement documentation.
Suggested solutions:
1. Check data availability
2. Decrease the prediction goal timeframe
3. If an Eligible population definition is provided, decrease the eligibility filter timeframe
4. If an Eligible population definition is not provided, try adding one (Error code: VALIDATION-407 NOT_ENOUGH_HISTORICAL_EVENT_DATA)
We require 120 days of recent data. For more information, please check the data requirement documentation.
Suggested solutions:
1. Check data availability.
2. Decrease the prediction goal timeframe.
3. If an Eligible population definition is provided, decrease the eligibility filter timeframe.
4. If an Eligible population definition is not provided, try adding one. (Error code: VALIDATION-407 NOT_ENOUGH_HISTORICAL_EVENT_DATA)
Suggested solutions:
1. Check data availability
2. Modify the prediction goal definition (Error code: VALIDATION-409 NO_OBJECTIVE)
Suggested solutions:
1. Check data availability.
2. Modify the prediction goal definition. (Error code: VALIDATION-409 NO_OBJECTIVE)
Suggested solutions:
1. Check data availability
2. If an Eligible population definition is provided, modify the condition or increase the eligibility filter timeframe (Error code: VALIDATION-410 NO_POPULATION)
Suggested solutions:
1. Check data availability.
2. If an Eligible population definition is provided, modify the condition or increase the eligibility filter timeframe. (Error code: VALIDATION-410 NO_POPULATION)
Suggested solutions:
1. Modify the prediction goal definition
2. Verify data completeness or use a different that includes examples of non-qualifying events for the prediction goal (Error code: VALIDATION-413 SINGLE_VALUE_IN_OBJECTIVE)
Suggested solutions:
1. Modify the prediction goal definition.
2. Verify data completeness or use a different that includes examples of non-qualifying events for the prediction goal. (Error code: VALIDATION-413 SINGLE_VALUE_IN_OBJECTIVE)