Tourism Media Automates Travel Content Generation with Neo4j AuraDB on AWS for an 8x Increase in Productivity
Tourism Media generates global travel content, but relational databases couldn’t keep up with changing model and scale requirements. With Neo4j AuraDB on Amazon Web Services (AWS), they could rapidly model geographic data for over 300,000 cities in the cloud, increase context, and automate content generation at a massive scale.
By the Numbers: Tourism Media’s Graph in Action
- Size of Graph: 7 Million Nodes; 300 Million Relationships
- Geographic Data for More than 300,000 Cities
- Using Graphs More than 100% Increase in Build Speeds
- Platform: Neo4j AuraDB Enterprise on AWS
Tourism Media is a small business with an in-house team and an extended collection of content creators across the globe. Seeking to bring better travel content to the online space, Tourism Media offers diverse services to their clients, from videos, photography, and content to entire content management and delivery systems. The company’s clientele includes travel industry giants, like Expedia.
The Challenge of Highly Unstructured Geographical Data
Tourism Media produces dynamic travel content for hundreds of thousands of destinations, but each piece of travel content needs to be customized and refreshed regularly for both geographical and seasonal reasons.
When it comes to data, geography is highly unstructured and difficult to model. Relational data models, with predefined, columns-and-rows structure, made it hard, and sometimes slow, to navigate data or produce relevant, meaningful content – like mapping out suggestions for road trips, translating local content into multiple languages, and automatically serving up top rated local hotels – in a scalable way.
“If we’re generating content around a ragged hierarchy like geography, how do we do that in a way that makes sense and scales?” said Terry Franklin, a freelance consultant specializing in Neo4j who worked with Tourism Media on its implementation. “For example, if we produce photos for the country Singapore, how do we make sure that they also appear on pages that understand Singapore to be a city?”
Tourism Media faced two challenges. First, they needed a way to ask complex questions of data that is highly connected and hierarchical without lots of JOINs and long-running queries. Next, they needed to ensure any new system they built aligned with their “Digital Transformation Roadmap,” which involved several key initiatives, including a move to AWS.
Graphs to Navigate Increasingly Complex Geographical DataThe Tourism Media team came across graph databases, and quickly identified their challenges as “graph problems” due to the nature of their geographic business with many points of data and relationship dependencies. They built a graph data model in Neo4j to determine how well it reflected their geographical data.
“Graph databases immediately became a much better fit, because they store your data in a way that human beings can intuitively relate to it much more readily,” Franklin said.
As their use case began to scale, the Tourism Media team needed to expand its adoption of graph technology. Neo4j AuraDB on AWS, with its pay-as-you-go pricing model and elastic scalability, was perfect for the team’s needs, allowing them to easily conceptualize and query data with increasing complexity.
Neo4j AuraDB on AWS – “A No-Brainer”
Once Tourism Media migrated their data to the AWS Cloud, it was quickly ready for production. Matthew Cassidy, CEO of Tourism Media called the move to AWS “a no-brainer” as it has become an industry standard, has an army of practitioners familiar with it around the world, and is already in use by many of their clients. It was also a key part of their digital transformation journey.
The Tourism Media graph in Neo4j AuraDB on AWS is flexible and expanding, with new data sources added to enable new products. Standard content templates use aggregated data coupled with simple, repeatable queries to create content for hundreds of thousands of destinations and millions of landing pages. Relationships map geographical hierarchies and allow context to be added on top of each template.
As Neo4j is a native graph database that continuously captures and directly stores these types of relationships and dependencies between data, it eliminates the need for additional processing, making it highly efficient with significantly less resource requirements.
Ontology-based weighting is applied to tags to provide relevant recommendations for similar places. Using only GraphQL, Tourism Media can now ask complicated questions of their knowledge graph.
“Neo4j, GRANDstack, and APOC really fit well together, because they allowed us to import all the data into the graph in a way that truly represents geography and ask the questions in GraphQL in the exact same shape that we expect the answer back,” Franklin said.
Tourism Media Architecture>
Build Times Reduced from Eight Days to One
Tourism Media’s knowledge graph now handles highly complex queries at a much faster rate. As they continue to add new data sources, Neo4j AuraDB on AWS enables them to scale their business and meet the needs of their clients, while reducing build times from 8 days to a single day.
“With over 300,000 cities in the system, some of our content builds took the relational database seven or eight days to complete,” said Franklin. “When we switched to a graph database, it took less than a day.”
Responses are “very positive from the clients,” according to Franklin. Reports show that content that is automatically generated performs as well as human-written content in terms of SEO – in some cases, even better.
The flexibility of the graph data model enables Tourism Media to move quickly to add new features including accommodation options, activities, attractions and flights – including whether it’s a direct flight or layover, how long it takes to get from a city to the destination, which planes are available, and more. They attribute an increase in quality of data output to using Neo4j AuraDB.
Tourism Media chose Neo4j AuraDB on AWS for multiple reasons. “It’s the reliability, the confidence that it’s always available and the fact that it’s managed,” said Franklin. “There is a good support team behind it, and it’s continuing to move forward. That makes AuraDB very compelling.”