How Data Science Is Predicting Medicine Shortages Before They Happen
“I Never Thought I’d Be an Accessory to Saving Lives”
Diana Okado didn’t set out to work in healthcare. Neither did her team.
Revolution Analytics is a Kenya-based data science firm, sharp on models, pipelines, and prediction. As Diana put it during a recent Impact Partnership session: “I was definitely never going to be a medical doctor. And I had not thought that by doing what I do, I would be an accessory to saving lives.”
That sentence is worth sitting with, because it captures something real about this project. The people who built it weren’t healthcare specialists brought in to digitize a known problem. They were outsiders who had to learn, fast, what actually mattered on the ground, and then build something technical enough to matter.
The Problem: A Blind Spot With No Warning
In neonatal units across Uganda and the DRC, clinicians face a heartbreaking moment more often than anyone would like to admit: reaching for an essential medicine and finding the shelf empty. Not because anyone was careless, because no one could see the shortage coming soon enough to act.
The team’s partner on the ground was Global Strategies for HIV Prevention. Nurses in their neonatal units use an app called NoviGuide to calculate medicine dosages for sick newborns, a tool already woven into daily practice, used roughly 300,000 times a year across 120 health sites.
What NoviGuide couldn’t do was look forward. Nurses only learned a medicine had run out at the exact moment they reached for it. By then, the gap between “running low” and “already gone” had quietly closed, with no warning in between.
That’s the real question that started this project, one Revolution Analytics, NoviGuide, and Tech to the Rescue asked together: could the data already being generated tell us when supplies were about to run low, before the shelves actually went empty?
Why Timing Mattered More Than Accuracy
Joy Ngugi, the data scientist who led the build, described how working in healthcare pushed her team to rethink their usual priorities. In most data science work, accuracy is the north star, get the prediction right. Here, that wasn’t enough on its own. The harder problem was timing: surfacing the right alert, to the right person, at a moment where it could still change what happened next.
That reframing shaped the entire system. Every time a nurse opened NoviGuide to check a dosage or follow a care protocol, the app quietly generated a signal, usage rates, delayed restocks, small fluctuations in consumption. None of these signals meant much in isolation. Fed into a prediction engine, together they became an early warning system, capable of flagging a likely stockout up to seven days before it happens.
Seven days doesn’t sound dramatic on paper. But in a neonatal unit with no local stockpile and a slow resupply chain, seven days is often the entire difference between reordering in time and running out. Behind the scenes, it’s cutting-edge machine learning. On the ground, it’s simple: clinicians get time to act.
Built Into the Workflow, Not Bolted On
The design decision that made this usable wasn’t just the model itself, it was where the model lives. The team built the first real-time data pipelines and is now moving into model training and dashboard integration, so predictive alerts appear directly inside the NoviGuide dashboards nurses already use.
No new app to download. No separate login. No added step in an already demanding shift, just an existing, trusted tool that quietly got smarter. That constraint, fit into the current workflow instead of adding to it, is often the difference between a technically impressive system and one that actually gets used in a busy ward at 2 a.m.
What “Impact Partnership” Actually Means Here
Global Strategies didn’t come looking for engineers who could simply write code. They said explicitly that they wanted more than technical skill, they wanted partners who cared about the work itself. That’s a harder thing to vet for than a portfolio, and it’s part of what the matching process was built to surface, giving both sides real confidence before the work even began.
The exchange runs both ways. Global Strategies gets a prediction engine protecting newborns across 120 health sites. Revolution Analytics gets something less measurable but just as real: a team of data scientists who now understand, firsthand, what it means to build for stakes this high, domain expertise they’ll carry into every project after this one.
What’s Actually at Stake
Every alert this system generates represents something concrete: a newborn’s chance at life. When medicine arrives on time, it isn’t luck, it’s insight meeting compassion, data doing what data should do, giving people enough warning to act with care instead of scrambling through a crisis.
None of this happens without the people behind it: the dedication of NoviGuide’s team, the support of Tech to the Rescue, the clinicians who turn data into action every day, and a data science team willing to be pushed outside their comfort zone. Together, they’re building toward a future where no baby waits for medicine that should already be there.
Want to see how predictive data could strengthen your work?
Visit us at revolution-analytics.co.ke or reach out at info@revolution-analytics.co.ke.