Azure Machine Learning based “Cross-sell Product Recommendations” available as preview in Dynamics CRM Online
Back in May 2016, Microsoft released a new feature “Cross-sell Product Recommendations” as a preview in Microsoft Dynamics CRM Online Update 1.
This Azure Recommendation Service looks back into the past transactions and data from opportunity, orders and invoices data within Dynamics CRM and analyse the sales trend for each customer – proposing you cross sell products. **Currently (last checked on June 12 2016), this function is a preview.
Until now, in order to understand the potential cross sell products for the customer, salesperson required some time to gain experience and also spending a lot of time on analyzing the past data. However, with this new feature in Microsoft Dynamics CRM Online, it will allow even the new salesperson to have a better chance of a successful cross sell.
In this article, I will be using the “Recommendation API” from Azure Machine Learning and integrate it with Dynamics CRM Online.
Prepare Microsoft Azure Machine Learning
Before you can setup Dynamics CRM, you need to have Azure Machine Learning prepared.
You do not have to create a Recommendation model from scratch as Microsoft has published a Recommendations API.
First, visit the link below to access “Microsoft Azure Marketplace”.
You can use this API for up to 10000 API calls for free. Click on “Sign up”.
Confirmation page is shown. Click on “Sign up” again.
Now the sign up is complete. Go to “My Account”.
In the right side of your screen, click on “My Data” to see the screen below. Click on “Use” next to the Recommendations API.
If you click on the Primary Account Key, a key is displayed. Make sure you take note of this key to use it later.
Prepare Microsoft Dynamics CRM for Cross-sell Product Recommendations Preview
Next is the setup for Dynamics CRM. The “Cross-sell Product Recommendations” function is still a preview so it is not enabled by default. It may be even possible that the instance you are using does not have the preview yet.
Go to “Settings”, and inside “Management”, access “System Settings”
Select the “Preview” tab, and next to the “Cross-sell Product Recommendations Preview” select “Yes” and click “OK”.
By enabling the preview, you should now have the option Azure Machine Learning Recommendation Service Configuration displayed. Select the link.
Disclaimer is shown. Click on “Continue”.
Enter the Primary Account Key you had issued in Azure Recommendation API into the Azure Account Key field.
Once filled, click on “Activate”
A new window is shown. Click on “Activate” again.
Now, your Azure Machine Learning and Dynamics CRM is connected.
Next, we need to have Azure Machine Learning learn the data from Dynamics CRM.
Go “Settings” and click on “Product Catalog”.
Click on “Product Recommendations”.
By default, “Recommendation Model” is shown.
Click on “Build Model Version”.
By default, it will create “Version 1”. Click OK.
Now, Azure Machine Learning has started to learn the data from Opportunities, Sales Orders and Quotations stored in Dynamics CRM.
Once machine learning is complete, the “AzureModelBuildStatus” should show “Success”.
Next, we will test out the model. Click on “Test Recommendations”.
“Test Model Versions” window is shown. Here, you can test out the product to be recommended on, and the model versions you have build.
Here, we will select the product below and the newly built model, “Version 1”.
Click on “Show Results”, the recommended products are shown as “Product Recommendations”.
Once the test are complete, setup the “Recommendation Version”,
and click on “Activate”.
Click on “Activate” again to confirm.
Try out the Product Recommendation
Now that the function is activated, we can try this out.
Go to “Sales” and click on “Opportunities”.
Select any of the existing sales opportunity, and scroll down to “Product Line Items”.
There should be a new column “Suggestions” and click on the “Suggestion” link.
A new window is shown, and the “Cross sell” recommended items are listed up. The “suggestions” tab have number between 0 to 1, and the higher number show a stronger suggestion.
If you want to add the suggested product, click on “Pick”
Now the recommended item is added to the Product line items.