Intelligent Recommendations account FAQ
This article provides answers to common questions or issues that might arise while using Intelligent Recommendations.
Contents
- Account Maintenance
- How do I create an Intelligent Recommendations account?
- How do I create modeling and serving resources?
- How do I contact the support team?
- Where can I find the serving point URL?
- My data contract has errors, what do they mean?
- How can I view the error logs for my modeling instance?
- Why can't I see any logs written to my logs folder?
- What happens if I misconfigure the date timestamp?
- My Model.JSON files/CSVs aren't being read correctly?
- What is the recommended approach to update my input data properly?
- API results and Error logging
- How can I check if I'm an authorized user for making API requests?
- The request has failed, but the URL is correct?
- I just created an account, but I'm not seeing results?
- How can I track the status of a model in my IR account?
- Where are my logging files or why are they empty?
- Why do only certain products or certain lists not return results?
- What should I do if my trending results are empty?
- What should I do if my "picks for you" results are empty?
- Why am I not seeing similar looks recommendations?
- Where can I find the full list of API status codes?
- Which parameters does Intelligent Recommendation supports for tracking API requests?
Account maintenance
This section is for questions related to configuring or maintaining Intelligent Recommendations accounts.
How do I create an Intelligent Recommendations account?
For steps on how to configure your account, go to Configuration overview.
How do I create modeling and serving resources?
For steps on how to create modeling and serving resources, see step 5 of the account creation quickstart guide.
How do I contact the Intelligent Recommendations support team?
For questions or issues not answered, you can reach the Intelligent Recommendations support team by opening a support ticket in Microsoft Azure portal.
- On the left menu, select Help + support.
- Select Create a support request.
- Complete the form. A support ticket is created.
Where can I find the serving endpoint URL?
For steps on how to locate the Serving Endpoint URL, go to your Intelligent Recommendations account on the Azure portal, and search for the Endpoint Resource you want to use. From there, you can copy the serving URL endpoint. For more information, go to the quick start guide for viewing recommendations results.
How to solve errors in the data contract?
To understand the error logs being returned in the data contract, go to Types of errors.
How can I view the error logs for my modeling instance?
To view the error logs for your modeling instance, set up your account to enable error modeling. For more information, go to Set up error logging. For more information about the types of errors that can be returned, go to Types of errors.
Why can't I see any logs written to my logs folder?
If you are not able to see any logs written to your dedicated logs folder, you can check whether your IR account has access to write to the relevant container:
- Go to your Intelligent Recommendations account.
- Select the Reports tab.
- Click on 'Check Access' button
- If the returned message says that 'The configuration for access to the storage account is missing', you need to grant writing permissions to access the associated container. If the returned message says that 'The requested storage account does not exist or you do not have permissions', please ensure that the given URL is correct.
What happens if I misconfigure the date timestamp?
If you're running into issues with your timestamps, the product is removed from the list of available candidates and filtered out from recommendations. To fix this issue, start by checking the ItemsAndVariants data entity inputs and ensure that the release date is in the supported date-time format (which is “yyyy-MM-ddTHH: mm:ss.fffZ”).
Issues reading the model.json CSVs
Opening the CSVs in Excel or any other non-CSV-friendly tool may remove empty columns that are required in the CSV format for Data Entities. We suggest opening and checking your CSVs in Notepad or a related app that won't modify the content.
What is the recommended approach to update my input data properly?
If you attempt to both model and update your input data at the same time, it can sometimes lead to modeling of data from mixed data versions and undesired recommendation results. See our Best practices Guide for more information about how to update your input data properly.
API results and error logging
This section is for questions related to API results or error logging documentation.
How can I check if I'm an authorized user for making API requests?
Users must be added to the list of authorized users for an endpoint resource before they can successfully call the service. For more information on how to check the status of authorized users, go to step 2 of the quick start guide for constructing API requests.
The request has failed, but the URL is correct.
If the URL is correct, then it's possible that the user who's making the request isn't on the list of authorized users for this endpoint resource.
For more information on how to check the status of authorized users, go to step 2 of the quick start guide for constructing API requests.
I just created an account, but I'm not seeing results.
If it just started, the service takes some time before results are available. You can track the modeling status for each of the models you've created. For more information, see our Modeling Status Reports Guide.
You can also look at the logs folder to learn about any issues with your data contract and whether you need to make any changes or corrections.
If you're still unable to see any results, ensure that you have given your Intelligent Recommendations account a permission to read data from the relevant container:
- Go to your Intelligent Recommendations account.
- Select the 'Modeling' tab.
- Choose the relevant modeling component.
- Select the 'Connection String' tab.
- Click on 'Check Access' button.
- If the returned message says that 'The configuration for access to the storage account is missing', you need to grant reading permissions to access the associated container. If the returned message says that 'The requested storage account does not exist or you do not have permissions', please ensure that the given URL is correct.
Also, verify that the correct folders and files are placed in the storage account. For more information, see Download the model.json file and configure the root folder.
How can I track the modeling status?
Intelligent Recommendations customers can track the modeling status for each of the models they've created on their accounts. After you configure a model, the service will periodically create a status log file to report on the current status of all algorithms (for your modeling tier). To learn more about how to access these logs, see Modeling Status Reports Guide.
Where are my logging files or why are they empty?
Your logging files might be empty if you didn't grant the correct Storage Blob Data Contributor permissions or pointed to the incorrect connection string. To confirm, follow our quick start guide.
Also, ensure that your error logging files have been configured correctly and verify that your connection string for your modeling instance points to the folder that contains your logging files.
Why do only certain products or certain lists not return results?
In general, insufficient interactions for a given item or user can affect the quality of results or the lack of any results being returned. Sometimes new products also need some time to acquire signals before showing up in certain recommendation lists. You can always look closer at your product performance using the dashboards.
What should I do if my trending results are empty?
The trending list API takes into account both the age of a product and its interactions count when scoring (ranking) each item in the list.
First count the number of interactions in a set time window, then our model factors in the age of the product (days since its release date as it appears in the catalog). The newer an item is, the higher the score, and as a product “ages”, its score decays exponentially.
To control the results in Trending, you'll need to adjust two configurable fields in the Reco Configuration data entity folder:
- TrendingListMaxAgeDays sets the maximum age in days of a product. A product that was released more than TrendingListMaxAgeDays days ago won't appear in the trending list.
- TrendingListTransactionsIntervalDays sets the time window in which interactions are aggregated. An item can be older than this value, but only interactions less than TrendingListTransactionsIntervalDays days ago is counted.
Adjust these settings in your Reco Configuration data entity and republish your model. Then you should start seeing recommendations results.
What should I do if my "picks for you" results are empty?
The "picks for you" capability is only available for modeling instances with a feature set marked as Standard or Premium. If you've confirmed that the feature set was configured correctly, take a closer look at the users you're trying to get recommendations for. A user ID without historical interaction data (such as previous purchases) won't return results when calling "picks for you". Also, confirm whether the user has recently opted out of receiving personalized recommendations. Any user who has opted out of personalization won't receive personalized recommendations.
Why am I not seeing Similar Looks recommendations?
Similar Looks recommendations are based on similar images and these recommendations will only be shown to customers who have selected the Premium feature set and have shared photos of their items in the data contract. Confirm that you've selected the Premium feature set for your model and configured images correctly as outlined in Image to item mapping data entities.
Where can I find the full list of API status codes?
The full list of API Status Codes is available on the Sample API Requests page, here.
Which parameters does Intelligent Recommendations support for tracking API requests?
It's optional to add one or more of the following tracking IDs as header parameters to your API requests. In case you encounter issues with your API requests, by adding one or more of those parameters we can track and investigate your requests more effectively according to the relevant IDs that you provide.
| Header key parameter | description |
|---|---|
| CorrelationId | |
| trace-id | |
| parent-id | Can be used according to your needs to group requests, for example of the same user, region, etc. |
Those header parameters can be added to any request type, regardless the URI, query parameters, scenario, etc.
See also
Quick start guide: Set up and run Intelligent Recommendations with sample data Quick start guide: Make an authenticated API call API Status Codes Error Logs Configuration overview Data contract overview