When it comes to the products they sell, many brands are all about quality and letting customer feedback speak for itself. This is true for many of our clients, including a national retailer we work to provide website services. In fact, when you visit its site, you are greeted by a map of projects and reviews in your area.
Related Content: How To Use Salesforce Marketing Cloud Data for Sitecore Personalization
But after transitioning its consumer-generated content platform from Modern Survey to Medallia—and with thousands of reviews coming in every month—the brand recognized the inefficiencies in manually moderating its customer reviews. Our client decided to ask DEG to implement a machine-learning solution that would not only automate the processing of customer reviews, but also provide an administrative screen built into Sitecore for its employees to provide necessary intervention if needed.
Finding Balance With Machine Learning
We chose to utilize Azure Functions to orchestrate the processing of the consumer reviews. The serverless computing approach allowed us to focus on implementing the necessary business logic without worrying about servers and infrastructure. We also took advantage of Microsoft’s line of artificial intelligence/machine learning services called Cognitive Services; specifically the Content Moderator. This service allows us to send in a review for processing and receiving various levels of scoring that helps our client decide whether a review is appropriate enough to display on its website.
The brand’s internal team also needed a way for humans to intervene in the moderation process and provide any corrections to the machine-moderated reviews. For that, we built a custom React-powered interface inside of Sitecore for surfacing moderated reviews. We began by adding a custom button to the Launchpad. Not only is this great for quick and easy access, but it also allows security trimming to be put into place if only certain people should be able to access customer reviews.
Users are able to sort and navigate through all machine-moderated reviews, giving them the ability to mark reviews—that the machine did not catch—as inappropriate or vice versa.
After launching the application, the user is able to sort and navigate through all machine-moderated reviews, giving them the ability to mark reviews—that the machine did not catch—as inappropriate or vice versa. Simply select a review to view details such as the text, associated affiliate store, and submission date. From there, you can choose to override any and all moderation codes the system has placed on the specific review.
How it Works
Below is a high-level diagram of our end solution, along with the flow of data from one system to another. Let’s walk through how everything works step by step.
- Reviews are sent from Medallia to the brand’s Azure Data Warehouse. This database serves as the single source of truth for both moderated and unmoderated reviews.
- Azure Functions pulls down any new, unmoderated reviews.
- Azure Functions runs each review through a series of moderation steps to determine whether a review is appropriate for public display.
- Azure Functions receives back fully moderated reviews.
- Moderated reviews are saved back into the Azure Data Warehouse. Steps 2-5 are repeated on an adjustable, 24-hour schedule.
- Moderated reviews are pulled from the Azure Data Warehouse and surfaced within Sitecore for the brand’s staff to intervene and provide updates if necessary.
- Sitecore notifies Azure Functions of any updates made to moderated reviews.
- Updated moderated reviews are saved back into the Azure Data Warehouse.
Sitecore component responsible for displaying reviews on the website pulls down reviews marked as appropriate.
Attending Sitecore Symposium? Make sure to attend my session on October 9 at 11:45 a.m.
DEG’s machine learning solution is currently in final testing and will launch in the coming months. From there, we plan to provide additional enhancements to further reduce any need for human intervention. Want to talk about machine learning and Sitecore for your project? Check out our Sitecore partnership page or send us a message.