Posit Team
Bringing predictive models to life for better patient and donor outcomes
Company
CSL is a leading global biotechnology company that develops high-quality life-saving medicines and public health solutions to help people with life-threatening medical conditions.
Size:
32,000 employees
Industry:
Biotech
Technology used:
Posit Team (Posit Workbench, Posit Connect, Posit Package Manager), Shiny, Databricks
Apps/products/solutions delivered:
Operations forecasts and optimization, decision support tools, marketing campaign metrics, text analytics, and many others
Serving patients in 100+ countries
$14.8 billion USD in revenue in FY 2024
CSL generated approximately $14.8 billion USD in revenue in FY 2024, making it one of the top revenue-generating biotech companies globally
Plasma donations are needed to manufacture life-saving therapies for people with bleeding disorders, cardiac surgery, organ transplants, burn treatments, and other conditions.
To ensure a high volume of essential plasma, it’s important that the donations and manufacturing are running smoothly.
Ben Bucior and his team of data scientists at CSL are using Posit to build and share real-time advanced analytics that drive informed decision-making across operations, marketing, human resources, manufacturing and other parts of the business to provide the best possible experience in their plasma donation centers.
"My team uses living apps to utilize deeply learned domain knowledge from our partners, and then our stakeholders get to have these models and visualizations at their fingertips — to run different what-if scenarios and work quickly and more effectively to deliver impact."
Evolving forecasting for dynamic 'what if' scenarios
Many forecasting journeys start with straightforward questions like: “What will our total sales volume be next year? What about in the next five years?” Initially, ad hoc analyses are done, yielding a number that is often shared via email or in a PowerPoint presentation.
For CSL, as with many companies, evolving business needs demand a shift from basic descriptive analytics to more sophisticated predictive modeling. Questions move from, “What will our total sales be next year?” to “If X happens, what will be the impact on Y?”
At CSL, this shift led to new, strategic questions like, “Given these internal and external factors, where should we build the next plasma donation center? If we launch a new pilot program, which centers would have the biggest impact on plasma donations?”
Ben and his team built models and simulations to address these complex, forward-looking “what if” questions. However, they still needed a way for it to become part of business stakeholders’ day-to-day workflows.
This desire to push the boundaries of their forecasting capabilities by enabling operations leaders and innovation teams to run what-if scenarios themselves, led Ben to start exploring new possibilities about two years ago - leading him to Shiny and Posit.
Results
Ben’s team now uses a lot of different ways to deliver data science model outputs:
Metrics (when you just need a number)
Dashboards (when you need to drill-down into specific groups of numbers)
Model predictions (when you need to estimate a new value or something in the future)
Living apps (when you need shared, high-value data products to co-design with your stakeholders)
For Ben, one of the most impactful aspects of building living apps with Shiny is how it enables the data science team to transform abstract models into concrete, interactive tools for business stakeholders. The data science team and stakeholders can now work side-by-side with a user-friendly front-end, facilitating faster iterations and clearer communication.
This iterative approach helps build a common language, share domain knowledge and find practical solutions for challenges at hand. With quick, seamless access to Shiny applications hosted on Posit Connect, stakeholders gain influence over model decisions, while data scientists develop deeper partnerships and engagement.
One example application, shown below, helps CSL rank more than 300 plasma donation center locations across the United States. Using Posit Team, CSL has evolved from a single univariate view of centers in a table on paper to a self-service, multivariate simulation tool that allows end users to incorporate both internal and external factors affecting the rankings. This shift has empowered operations teams to pinpoint opportunities to improve the donor experience for increased plasma donations by identifying future locations and optimizing operations at existing centers.
Power in the flexibility of Shiny for bringing models to life
Collaboration with the business enabled the data science team to develop predefined scenarios in Shiny, supporting CSL leaders in decision-making for ranking plasma donation centers by factors like operational efficiency and center size.
The flexibility and customization possible with Shiny immediately stood out as a significant advantage over proprietary business intelligence (BI) tools. Code-first design in Shiny is amenable to best practices from software engineering (such as version control and automated testing) and is extensible using R and Python packages. When end users request additional features to improve their experience – such as a search box, a particular model visualization, or a pre-formatted download – they aren’t constrained by software limitations. Instead of reworking business problems to fit into a specific technology, the teams can focus on identifying the best business solution, then build the technology and user experience to bring that concept to life.
This adaptability facilitates faster problem-solving. While they may not know in advance what insights a particular data set will reveal, that flexibility and process of iterating enables them to more quickly get to the problem they’re trying to solve.
As stakeholders began using the tool, they were empowered to conduct their own exploratory analysis, often discovering insights like, “This one aspect we thought was very important actually isn’t helpful at all.”
Ben adds, “It’s amazing how quickly the business can change with that. With this tool now people are asking about the limits of possibility. Can we do this idea? Can I get access, I want to look at this? It’s really cool seeing that transformation in just 1.5-2 years from the very beginning of this app.”
The self-service simulation referenced above is one of many models now used to make business decisions. Their internal platform, built on Posit Connect, brings dozens of models and datasets together, allowing users to access the full range of descriptive analytics through predictive analytics.