Understanding before you build: strategy-first work with Jacaranda Health

Jacaranda Health is a maternal health organization headquartered in Kenya. Its flagship product, PROMPTS, is an SMS-based service that supports millions of mothers across East and West African countries with maternal and child health information. The heart of the service is powered by AI, where mothers can get personalized and context-aware answers powered by AI models that triage, classify, and respond to messages from mothers in real time. From the outside, it looks like a technology story.

But when Jacaranda's leadership asked us to help in 2025, we spent our first months deliberately not building anything. They reached out because they felt limited by their data. Their data was not reliable, and their understanding of the operations was brittle, despite having an in-house data team and an existing data warehouse. So we worked to understand the business first, and let the technology follow. That choice turned out to be the strategic heart of the engagement.

When the AI outruns its foundations

PROMPTS is a striking example of a pattern we see more and more. The visible product, the conversation a mother has with a nurse-backed AI, is the tip of the iceberg. Beneath it sits the data: message logs, support tickets, M&E surveys, and programme metrics. At Jacaranda, the AI layer had advanced faster than the foundations it depended on. Data lived in fragmented systems, with inconsistent quality, unclear ownership, and limited governance. For a service where mistakes carry real clinical consequences, that's a brittle position.

The instinct for many organisations, and honestly, for many consultants, would be to start building: a new model, a new dashboard, a new pipeline. But before we could recommend any of that, we had to understand what existed, what mattered, and what would actually move the needle. If you don't understand the business first, you can build the wrong system with great confidence.

That understanding structured our work around three pillars.

Assessment: diagnose before you design

We started upstream, with a diagnostic rather than a build. We mapped the data systems, the decisions people actually make, and the dependencies between them. What we found was a sophisticated organisation under strain: a fast, built-for-speed engineering culture that had outrun its own data management, confusion about what the data warehouse was really for, a sprawling web of custom ETL, and an open security posture that needed attention soon.

The most valuable finding was that many of Jacaranda's "data" problems were upstream problems. They were process and tooling problems, not modelling or metrics problems. Fixing those would do more than another dashboard ever could.

This is where strategy often lives: not in choosing a technology, but in working out the right order of operations, and what not to do. Once the priorities are clear, the technology can follow.

The data strategy: a target state

The diagnostic gave us the heart of the problem, but naming the problem is not the same as solving it. So we took the time to design a target state for Jacaranda: the technology, the people, and the processes that would be needed to ease the pain we had found and let the organisation reach the next level of operational scale.

This was not a technology-selection exercise. We produced a data strategy Jacaranda could actually execute, one that set out what to change, what to build, what to stop doing, and in what order. Every element was anchored to a real decision the team needed to make, which kept the plan grounded in the business rather than drifting into stack talk. Technology was part of it, but it sat alongside the team structure, the ownership of data, and the processes that would keep it all healthy. Getting this right is why the early wins that followed could compound, rather than stay one-off experiments.

Proof of concept: early wins

Assessment earns its keep only if it leads somewhere. So we followed the strategy with a deliberate first win: a small dbt prototype that proved the modern workflow approach on a narrow, real transformation. Momentum built from there: a custom connector for the messaging platform that untangled a knot of bespoke ETL, a master-data prototype that surfaced duplicates no one had previously managed, and, as the flagship proof, a first-country reporting suite delivered end-to-end on the new stack, from ingestion through transformation to a QA dashboard people could actually use.

These weren't big builds. That was the point. Fast, concrete wins showed the direction could work, earned the team's trust early, and gave Jacaranda's leadership a credible story for why the modernisation deserved budget.

Leadership buy-in to get full understanding

Technology has to always fit within the wider organization, and any change needs to be understood by the right stakeholders. For this work, we invested time to share our recommendations with Heads and Directors, aligned leadership on priorities and sequencing, and built the capacity plan together with Jacaranda's data team. The goal from the start was a system the internal team owns and evolves, not one that depends on us to run.

That commitment to leadership alignment and knowledge transfer is exactly what turned good work into a partnership. When Jacaranda's leadership trusts the diagnosis and understands the direction, the work stops being a vendor relationship and starts being a shared project with the good of the organization in mind.

Where 2025 left us

Jacaranda closed the first phase of the engagement in December 2025 with one country on the path to the new approach, an agreed target architecture, and a data strategy Jacaranda's own leadership had genuinely bought into. The assessment had identified the problem, the strategy set the direction, the early wins had proven it, and the team was in a position to own what came next.

"Ona Insights has proven to be an exceptional and highly professional data partner, and I wholeheartedly recommend them for organizations seeking expertise in data architecture, technology strategy, and infrastructure improvement."

— Anneka Wickramanayake, Director, Research, Design and Evaluation, Jacaranda Health

Ona Insights is a Nairobi-based data and AI partner for impact-driven organisations working in complex environments. We make sense before we build, and we build so our clients own what we create.