A product recommender system for the housing market
Compared to a few decades ago, purchasing a home today is more of a digital experience. Home buyers have easy access to online educational tools and resources, making them hands-on and proactive with their home search. Customer choice is overwhelming in today’s competitive environment. In addition, the amount of information that customers encounter daily makes selecting a product or service difficult. In this article, I will share our journey in customizing Spinor AI for our customers in the housing market.
Our solution: Personalized recommendations for home buyers
Our team of experts created Spinor, an AI-powered recommendation engine designed to drive customer value through a rich and relevant multi-channel experience using AI predictions. As part of our solution, we use the Force.com platform’s computation capabilities, customer databases, and product information systems. To provide personalized product recommendations to customers, we use Salesforce Marketing Cloud and CMS.
Boost your revenue and drive customer value with an AI-powered recommendation engine.
Our customer needed us to create personalized recommendations for their home buyers. This required our system to propose products like apartments or houses to their buyers. These recommendations needed to be individually personalized using first party data and shared through email and integrated with their website feed. With our system, we are able to provide our customer’s buyers access to individually personalized domicile home recommendations, and collect buyers insight and measure their engagement for our customer.
Analysis of the proper AI recommendation model
The most important initial step in any product recommendation project is to analyze the data available and needed for buyers and products. After this step, we analyzed the best algorithm for the task. Once we completed the analysis, we found that the correct class of algorithm is a feature-based product recommendation. To achieve data-driven personalized marketing and to implement feature-based capabilities, we needed to enhance Salesforce’s collaborative filtering algorithm. An enormous benefit of the selected algorithm is the lack of the “cold start” problem. We can start providing recommendations the moment we have the first data in the buyer’s behavioral interest profile.
First area we needed to design was customer behavioral interest profile with feature-based activity scoring model. Second area was to design and map the data of the products database to select the most relevant features of the domiciles that are core decision factors for buyers.
Let’s get started: Customizing a recommendation model in Spinor AI
For customizations of Spinor AI product recommendations, we divided our efforts into the following work streams:
- Business requirements
- Customer Data Model
- Products Data model
- System integrations
- AI recommendation algorithm customisation
- Machine Learning process
- UI component for Salesforce sales users
- Marketing Cloud journey definition and email templates
- KPI reporting for recommendation effectiveness
Let’s check it out: Testing the solution
As the solution was ready for testing, we started end to end comprehensive testing simulating various use cases. The amount of different product recommendation test cases was almost 100.
Let’s make it run: Designing an operational architecture
Now it was time to design the way the solution will be monitored and maintained. The nature of the AI empowered solution is its constant improvement. For our case we needed to make sure that we are being relevant in scoring model for buyers actions, proper products features were selected and how well every recommendation resonates with individually for buyers.
Last but not least: Finalizing a deployment strategy
As the AI solution was a new experience for our customer we deployed the solution using a geographical model. First, we started in one country and after fixing all initial problems we will deploy for the entire organization in all remaining territories.
Continuous improvement: A/B Testing
Personalized product recommendations are AI estimated predictions for buyers needs. Once the recommendation system was deployed to production, A/B testing was conducted to gauge its effectiveness by metrics like click-through rate (CTR) and engagement level with and without the recommendations. Metrics data are used to tune the AI algorithm strategy as well as optimize the customer journey.
In Summary
I hope this article provided some food for thought for developing and deploying Spinor AI – personalized product recommendation engine. Looking at a project results holistically, with buyers’ convenience for shortening the domicile selection process time in mind. The solution delivers real measurable business value and has a positive impact on how the project is perceived both internally and externally.
Are you ready to take your customer experiences to the next level?
Genus One has a team of dynamic IT professionals that provides a complete range of Salesforce professional services under one roof. Our flexible approach allows us to provide a wide range of support models to suit all business types.
We aim for superior performance across all aspects of your Salesforce business operations. Our Salesforce support team knows how to identify problems & challenges and address them head-on. We can take care of everything smoothly and smartly, ensuring the best results!
