Real-time recommendations have taken their place as a crucial part of today’s online experience.
After starting in the online dating and e-commerce sectors, recommendations are now being used in all industries as merchants and service providers rush to provide personalized, easy-to-use services to online user communities.
Recommendations Change the Rules
The key to success in the recommendations game is producing instant, relevant suggestions to customers while they browse your website. While this might sound simple, the associated technical hurdles are massive:
- Your recommendations must be personal and not just a list of popular offerings
- You must analyze user preferences, browsing histories, demographics, orders, products, categories, prices, and even other users’ habits and buying patterns
- You must consider up-to-the-moment browsing data that was gathered seconds ago
- Suggestions must be produced with sub-second response times
These requirements call for special technology, namely a graph database. Early Internet leaders—such as Google, Facebook and LinkedIn—had to write their own graph recommendation engines, But Neo4j now offers a commercially-ready graph platform that you can use to make real-time recommendations to your customers.
Relationships Drive Recommendations
Neo4j can connect massive amounts of connected data in milliseconds. And its gives equal importance to data elements—such as customers, products and orders—and the relationships between them: who likes what, who browsed which products, who bought which products, which purchase happened first, who is similar to the current user, etc.
Neo4j also supports any number of relationships between entities as well as the quality and strength of each connection, so you can create semantically-rich data ripe for making recommendations. In addition, Neo4j can make recommendations in both directions, so you can identify items of likely interest to individuals, as well as find individuals likely to be interested in a given item.
The Need for Speed
To maintain the interest of your site visitors, you must be able to provide relevant recommendations almost instantly. This is another area where Neo4j’s graph technology shines.
Neo4j easily outperforms traditional relational and NoSQL databases when connecting massive amounts of buyer and product data. For example, in milliseconds Neo4j can find users who are single, know each other, who own and like a specific game console, and repeatedly visit a specific website.
Such a query using traditional models would take days to write and potentially many minutes or hours to execute. In contrast, with Neo4j such queries are easy to specify, and super-fast to execute since there are no speed penalties from slow, expensive and unpredictable table joins.
So by using Neo4j, you can instantly provide relevant recommendations—to keep customers on your website and close more business.
“We found Neo4j to be literally thousands of times faster than our prior MySQL solution, with queries that require 10 to 100 times less code. Today, Neo4j provides eBay with functionality that was previously impossible.”
Volker Pacher, eBay senior developer
Recommendations Raise Tough Data Challenges
To create meaningful recommendations, you need to amass and process large volumes of data and relationships. As application grows, you must be able to handle an explosion of data elements and their complex, interconnected relationships—while maintaining insanely fast query response times.
Neo4j’s graph technology is designed to meet the demands of the rapidly and continuously growing datasets created by real-time recommendation applications. Its ability to rapidly traverse connections in highly connected data enables you to efficiently store your ever-growing database as well as maintain rapid response times.
Business Benefits of Neo4j
Organizations across a wide variety of industries worldwide have selected Neo4j graph database as a core platform for business-critical systems that drive revenues and improve customer service and satisfaction.
The primary benefit of using Neo4j as your recommendation engine is to produce relevant recommendations that appeal to customers. Neo4j’s ability to consider all relevant relationships and user histories in real-time maximizes the accuracy of your recommendations. eBay cites that “Neo4j allowed us to add functionality that was previously not possible.”
Neo4j’s graph query engine delivers fast recommendations for even the largest query loads and the most complex network of users, products and buying patterns. Telenor, one of the world’s top telecom providers, switched to Neo4j on its business customer portal and recognized a thousand-fold performance improvement.
Higher Transaction Sizes
Neo4j can suggest product recommendations, upsells, cross-sells and alternatives for out-of-stock products to raise average transaction size across shopping carts and raise overall revenues.
Faster Time to Market
Graph applications built in Neo4j take a lot less code and a lot less time to build than traditional relational and NoSQL applications. eBay’s Neo4j recommendation apps required 10- to 100-fold fewer lines of code than their SQL-based predecessors. And less code translates directly to faster time to market.
Easy Maintenance and Enhancement
Neo4j graph datasets can grow easily, even when you add new data sources and when data structures inevitably change. While traditional technologies require you to redefine data schemas and rewrite backend code when such changes occur, Neo4j’s schema-less data model allows you to flexibly maintain and enhance your recommendation engine.
The efficiency and scalability of Neo4j allows you to deploy applications with 10 times less hardware than traditional database solutions. Its clustered, fault-tolerant server architecture allows you to add and load-balance servers as needed, and its multi-data-center support lets you deploy applications on a global basis.