When the NRL’s tiny data team scrummed to find a way to help the company make smarter business decisions faster, they scored more than they expected. Even with the final whistle yet to blow, the project is looking like a shrewd product play.
As the wholly-owned corporation set up by the non-profit governing body for the sport from which it takes its name, the Australian Rugby League Commission (ALRC), “NRL” is one of the most widely understood and affectionately regarded acronyms in households across the country.
The NRL has close to six million members and runs 204 games each season in competitions involving 29 clubs across Australia and New Zealand. That includes the annual interstate grudge match between NSW and Queensland, State of Origin.
The sporting body schedules matches at over 25 stadiums and smaller venues, manages multiple sponsorship agreements, and has ticketing partnerships with providers and betting agencies, such as entertainment booking giant Ticketek.
Its lineup of advertising, sponsorship and broadcast deals also keep it busy. Its main partner is Telstra, and it has media rights deals with Nine, Kayo and Foxtel. Its advertising sponsors include Ampol, Chemist Warehouse, Westpac, Youi, Kia, Swyftx, KFC and, controversially, wagering companies like Sportsbet.
It also has hundreds more business relationships ranging from official game merchandise vendors to the innumerable food and hospitality providers the code supports.
When it’s not attending to commercial matters, it’s cultivating an ambitious community engagement program that reaches right back into the grassroots of the sport, with junior division players across Australia starting out in the code from ages as young as six.
Digitally, it’s all held together and supported by more than 12 private data sources along with six other connected public data sets. The data flows constantly in real-time.
Arguably, the task of converting all that data into a form that could help any organisation make meaningful strategic decisions would make even a seasoned chief data officer a bit dizzy.
Lining up for the goals
In NRL’s case, the job fell to its head of data, Albert Xavier. Speaking at Databricks’ 2025 Sydney event last week, Xavier said that the NRL started lining up for the project last March.
The sporting body started using Databricks’ Lakehouse, Lakeflow Connect, Lakeflow Jobs and its Unity Catalog to integrate the NRLs siloed data sources into a cloud-based data lake. Just over eight months later it had a single source of truth with data flowing in real time, and was ready to execute its vision.
“We wanted to give power to people who can actually make the difference, the business. We did not want to just show them numbers. We wanted to tell a story behind it,” Xavier said, before reeling off just a sample of the kinds of decisions his team of data boffins hoped to support.
“We need to make sure that they could take care of the pricing strategies – How many products are in there? In which venue? How many seats are there? How many kill seats (nosebleed seats often heavily discounted or decorated to make a venue look full for the cameras) [are there]?
“Which business is going to sell what kind of a product? What is the hospitality going to sell? Is [the data] real-time? What is the marketing strategy going to look like?
“These are just 20 percent [of the questions]. Based on that we thought: ‘how are we going to solve this puzzle?’,” Xavier said.
Xavier’s team wanted to give the organisation the confidence that the data and the insights it provided were current.
“That’s what everybody expects – when business takes a decision, they want to be really confident of that – they want to take that decision on a real-time basis,” he said.
Ordinarily, this might mean taking on an extensive program of work, building a frontend presentation layer and integrating governance, machine learning and agentic AI piecemeal.
Xavier directly manages a team that has a total headcount of just three. He only calls on external resources for larger projects.
The internal team put their heads together, took stock of their current situation and looked for a way to execute the vision.
Points, but not the good kind
Xavier said that the NRL had been managing its data using a primitive method that relied on macros used across multiple Microsoft Excel sheets. The macros became sluggish as volumes grew, Xavier said.
Also, the Excel sheets weren’t automated and in turn relied on data analysts for accuracy. If the analysts weren’t available, the task fell back to the business and the chance of data corruption grew. If one of the NRL’s many partners asked for a specific data set, that too required a manual process to isolate and send in a digestible.
The NRL could have tried a notebook approach or a dashboard, but notebooks can be a bit unfriendly to business decision makers and dashboards limited in functionality.
Xavier’s team also needed lineage and governance controls, and to have the flexibility deploy analytics, agentic AI, genies and machine learning.
The NRL could have used multiple external teams to build a new frontend. However, after considering the need to draw up business requirements documents, attend to maintenance and lineage, link the frontend to the lake house with a caching database in between, Xavier said the project’s scope was gathering complexity rapidly.
The internal team had what Xavier on reflection now sees as a stroke of good fortune.
“The best part was that the team that we contacted to do the front end were busy for the next three-to-four months. That’s when we started to look at other opportunities in the market,” he explained.
Putting the (internal) team first
The team considered an emerging approach to the problem. Its centralisation vendor, Databricks, suggested using an emerging application-based concept it has been championing globally for the project.
The idea behind it is to bring the app to the data, rather data to the app. Generally, apps are built in a hosted environment and then their developers find ways to take the data from their lake houses to them, which, once governance, maintenance and lineage concerns are taken care of, can be expensive and complex.
The Databricks approach is to sit the application development environment directly on top of database infrastructure. Databricks would supply the team with Databricks SQL Warehouse and a few other authentication tools to get it started. After that it would provide them with a special URL and a string of code; the team could then build the app in whichever popular open-source framework they fancied.
The NRL’s internal team did proof-of-concept tests involving three to four tools, benchmarking them against its checklist for business goals.
It decided to pull the Databricks approach from the benches and give it a run in the paddock
Keeping score: 60 percent cheaper and 80 percent sooner
“In one single architecture, keeping [it] simple within Databricks. What did we achieve?” Xavier asked or the benefit of participants before answering “almost 60 percent of the cost was saved, and we finished almost 80 percent faster.
“I’ll tell you why – because we did it between just one single team. We did not actually involve the [external] teams which [we’d have needed otherwise].
“It just took minutes for us to create the apps. I would say, leaving the initial data loads, it took just six weeks for us to finish the project end-to-end,” Xavier said.
“If the business has got any problems, all they have to do is come and see us and we will fix it for them. It’s less dependent on any others to have a look at it,” he explained.
The NRL’s team chose Streamlit’s coding framework to build the application and Databricks’ Unity Catalog was already taking care of security concerns like governance and lineage.
The NRL is currently putting its first application, one to support its ticketing operations, into user acceptance testing with a schedule to send it live in February, before the next season kicks off.
It also plans to start integrating agentic AI into the apps.
Xavier said that the team achieved more than simple business intelligence project. They realised that many more business uses were possible just by slicing and dicing the data using filters under the app.
“You see, what we have created here?” he challenged the Databricks attendees to consider. “It’s like we started off trying to solve a problem for the business but then eventually we built a product,” he said.
He even went a bit further with the analogy.
“It started off like a technology project, but I think it was a transformation that we were doing”.
We’ll be right back after the break.