How BES makes health data useful – and the team that does it.
Every day, digital health systems built by BES help visualise thousands of data points across the Pacific and beyond. But data isn’t useful just because it exists. It has to be structured, validated, and presented in a way that helps people make real-world decisions. That’s the job of our data team.
In this post, we share a bit about how that happens – the people behind it, the processes that keep it running, and the work we do to turn raw data into something that matters. Our data team spans Melbourne, Sydney and Canberra, and brings together a mix of backgrounds across analytics, software engineering, and biomedical research.
Meet the team
Juliana Mitchell-Wong leads the team. With a background in biomedical and computer engineering, she brings both deep technical knowledge and practical experience leading the evolution of our data systems over the past four years.
Meet the team
Salman Hussain Raza first joined BES in our Fiji office and is now based in NSW. His background in software engineering means he has a sharp focus on automation and process design.
From point of care to point of impact
Data in Tupaia comes from many sources. Our own data collection tools — MediTrak on Android and DataTrak in the browser — are widely used, especially for surveys and field reporting. But we also pull data from other systems including Tamanu, DHIS2, mSupply, Kobo, and Google Drive, among others. These integrations are designed to be flexible: we can ingest data through APIs, database exports, backups, or spreadsheets, depending on the context.
But just ingesting data isn’t enough. Once we have it, we need to make sure it’s clean, consistent, and safe to report on. That’s where our processing pipelines come in.
Built-in checks and balances
Our data pipelines process over 30 million data points every single day. For that to work reliably, we use automated checks to detect issues in real time, before they impact reports.
When new data arrives, it goes through a series of steps that automatically:
- Remove any deleted, outdated, or otherwise unneeded records
- Check for issues like missing values, data integrity failures, etc.
- Flag inconsistencies for our team to review
If something looks off, it doesn’t just sit there quietly. A validation error is raised and the team is notified. These early alerts are often what allow us to catch issues before they make it into a report or dashboard. It also means we can go back to the original data source, and make sure it doesn’t happen again – whether that’s through user training, or requesting a bug fix in the source system.
Meet the team
Gagandeep Kaur, based in Canberra, brings a strong interest in how data is interpreted and communicated. She works closely with our product and implementation teams to ensure what we visualise is not just accurate, but also meaningful to the people using it.
Meet the team
Georgia Lack is our Head of Impact and Partnerships and has a public health background, rather than data engineering. She works on the ‘country’ side with our partners to understand what data they are collecting, what data outputs they need, and how they can use that data to improve decision making.
Making it easy to report on what matters
As data is ingested, validated, and cleaned, we simultaneously structure it into reporting-friendly database views. We also apply standard calculations at this database view layer, so that metrics are consistent from one visualisation to another. This helps ensure dashboards are fast and reliable, while also keeping the reporting logic clear and easy to follow. Tamanu reports use straightforward queries on these views, and the data from them flows through the rest of our data pipelines. Once in Tupaia’s Viz Builder, they can be combined, filtered, and visualised however needed.
A real-world example: forecasting medication shortages
In September 2023, the Samoan Ministry of Health carried out a mass drug administration for lymphatic filariasis and scabies. Doses were weight-based, and recorded each evening in Tamanu. These were visualised immediately in Tupaia.
After a few days, the dashboards showed something unexpected: stock levels were dropping faster than anticipated. Because of how we model and visualise data, the Ministry was able to quickly spot the trend and take action — placing an emergency order for more medication to ensure the campaign could continue.
That’s the kind of impact we want our data systems to have: helping people see the issue early, and act on it.
To read more about this case study and others like it, visit Data for decision making case studies from the Pacific.
Why it matters
Whether it’s a national outbreak dashboard, a forecasting tool, or a facility-level report, our goal is always the same: to make health data useful. Only then can it be used to improve patient safety, make better policy decisions, or strengthen public health programming.
It takes a solid system, a clear process, and a good team to get there. Fortunately, we have all three.
