Pay.com.au can now see its whole customer journey. It took 16 weeks.

At a Glance

• End-to-end customer visibility for the first time: onboarding, platform events, payments and marketing, correlated in one place
• Executives self-serve answers in plain English through Databricks Genie, without waiting on the dev team
• New data sources onboarded in days, not months
• Churn models and in-product recommendations now built on complete data rather than the slice the data team could reach
• 16 weeks from kick-off to production, across two regions
• One Datalynk consultant, working alongside a three-person Pay.com.au team

“It’s a game changer.”

— Paul Moreton, Head of Data, Pay.com.au

The Challenge: A Business Growing Faster Than Its View of Itself

Pay.com.au grew fast. Like most companies moving from start-up to scale-up, its data grew faster than anyone could control it.

Payment transactions lived in a custom-built payments system. Customer data lived in Salesforce. Bespoke Python processes and four or five other platforms each held their own slice, and each served up its own version of the truth. Wherever data crossed a department boundary, teams made a copy and shaped it by their own rules. Most of it ended up in spreadsheets.

The practical consequence for leadership: a cross-business question was close to impossible to answer. Data wasn’t accessible outside the development team, so an executive wanting to know how onboarding behaviour affected long-term customer value had to queue behind engineers, stitch together exports, or go without. Analysis could only ever go one department or one system deep.

Even a question as fundamental as “what is a customer?” had several competing answers: a business entity in one system, an individual user in another. The data team, meanwhile, was building machine learning models locally on the limited data they could reach, knowing the models were only as good as the slice they could see.

This wasn’t a talent problem. With a development team of more than 100, Pay.com.au had already invested in AWS Athena and Salesforce Data Cloud and evaluated plenty of other options. Point solutions couldn’t fix an architectural problem, and the team recognised that building a data platform is a specialist discipline in its own right. They needed deep, focused experience, and they needed it fast.

Datalynk came in on personal recommendation, backed by our credentials as a Databricks partner in Melbourne.

What Changed: From One System Deep to End-to-End

For the first time, Pay.com.au can see the complete customer journey. Platform events, payment transactions and marketing data are correlated end-to-end, from the moment someone starts onboarding through to their first payment and beyond. Analysis is no longer trapped inside departmental silos.

Executives answer their own questions

The capability leadership never had before, and by their own account the one they value most, is Databricks Genie: natural-language questions answered directly from governed, trusted data.

A concrete example. An executive can now ask how specific application events affect a customer’s likelihood of completing onboarding in full and going on to become a long-term customer. Before the platform, that question crossed three systems and a development queue; today it’s answered in a sentence, with no analyst in the middle.

Better models, better product

The data team that was building machine learning models locally, on whatever data it could reach, now builds on the complete picture. The result is significantly better churn models and very specific customer recommendations that are now live inside the Pay.com.au product. The business benefit isn’t a dashboard; it’s a better experience for customers and a sharper view for the executive team.

Speed to delivery

Onboarding a new data source now takes days, not months. Every future dataset follows the same documented pattern, so the platform compounds in value with each addition rather than needing to be re-thought each time. For a lean three-person data team, that shift in speed to delivery is the outcome that matters most.

Questions that took a week to answer are now answered in minutes, without the data team needing custom extracts.

“We couldn’t have done it without your help. You’ve delivered something we’ve never had at Pay.com.au… a repeatable data platform.”

— Paul Moreton, Head of Data, Pay.com.au

How It Was Delivered: One Platform, Two Regions, Built Once

Those outcomes rest on a foundation that was designed to be run by Pay.com.au’s own team, not by consultants. Over 16 weeks, a single Datalynk consultant worked hand-in-hand with three people at Pay.com.au, with architectural guidance from their side, to design and deliver a complete Databricks Lakehouse. It is deployed in both Australia and the USA to meet data residency requirements, and serves both markets from one consistent architecture.

Start with the business definitions, not the code

The hardest problems in this engagement weren’t technical; they were definitional. Before a line of code was written, we used a bus matrix to agree the structure of the business layer with Pay.com.au’s stakeholders, settling questions like “what is a customer?” once, for everyone. We then wireframed the dashboards so leadership could react to what they’d actually get before anything was built.

Understand the goalposts first. Build it right, build it once.

A medallion architecture built for a payments business

• Bronze: Raw data landing in micro-batches every 30 minutes, with full history preserved for every source, including the custom payment transaction system.
• Silver: A deduplicated, verified copy of each source: clean enough for machine learning, faithful enough to trust.
• Gold: A dimensional model that adds the business context, powering analytics and Databricks Genie, so executives can ask questions of the data in plain English.

The same dBT models drive bronze and silver in both regions; only the gold layer diverges where each market genuinely differs. One codebase, two countries, zero duplication of effort.

Serverless by design

We deliberately chose Databricks serverless. It meant trading away a handful of features available in self-hosted deployments, but in exchange Pay.com.au’s team never has to manage cluster configuration, capacity or infrastructure. For a lean data team, that trade-off pays for itself every week.

A framework, not just a platform

What Pay.com.au received wasn’t a one-off delivery; it was a repeatable way of working that their own team now operates:

• Git-controlled CI/CD pipelines for every deployment
• Hundreds of automated data quality and coding checks
• AI-assisted code review
• Standardised dBT project structures
• A common, documented pattern for onboarding every future dataset

The operating numbers

• Data refreshed every 30 minutes across every source, with full history retained
• 99.999% pipeline uptime
• One consistent architecture across the USA and Australia

These are the reasons the business outcomes above are durable rather than a one-off project high.

What’s Next

The foundation is built, and Pay.com.au’s roadmap is simple: more.

More source systems onboarding through the standardised framework, in days. More models graduating from local environments onto Databricks, with fraud detection next in line. More Genie in the hands of executives, and fewer questions that have to wait.

That’s the real measure of a zero-to-one platform: not what was delivered on day one, but how fast everything after it can move.

Growing from start-up to scale-up is messy. If your data is growing faster than your control of it, talk to Datalynk about building a platform your team can run, and build on, themselves. Fast delivery, long-term impact.