By Chidi Larry

An expert in data engineering and platform design, Emmanuel Ezekiel has shared practical insights on how Nigerian tech teams can build reliable data systems that thrive despite poor networks and limited budgets.

Drawing from his experience supporting finance and consumer applications, Ezekiel says every strong data platform begins with one rule: treat data like a product—with users, service levels, and costs.

“Architecture is a means, not the goal,” he explains. “Start with the decisions the business needs every week, then choose patterns that serve those decisions.”

He says successful projects begin with the right questions, not tools. Teams map out recurring business decisions—fraud detection within 15 minutes, daily net inflows by segment—before designing tables and models. These questions define structure long before any software choice is made.

Ezekiel advises against blanket rules for pipelines. ETL, he says, fits regulated or PII-heavy domains where transformations must occur in controlled spaces. ELT, on the other hand, suits agile environments where raw events land first, then transform with versioned SQL and published contracts. Each contract must clearly define schema, semantics, and ownership.

He insists that contract tests should fail loudly when definitions drift, while column-level lineage must let leaders trace KPIs to their origins. Every pipeline should have an assigned owner, clear freshness targets, and a recovery playbook. “When accountability is clear, reliability becomes routine work, not a nightly crisis,” he says.

Ezekiel also highlights cost as a design factor. Cold data should move to cheaper storage, unused transformations should be pruned quarterly, and vital dashboards should be cached. “Publish cost dashboards by team so everyone sees what they spend,” he advises. “The cheapest query is the one you don’t need to run.”

He believes these principles fit Nigeria’s realities—unstable networks, diverse devices, and tight budgets. Systems that succeed here, he says, are simple, transparent, and quick to recover. “Powerful tools are arriving,” Ezekiel notes. “If we keep our bias for clarity, we’ll build platforms that are not just impressive, but truly useful.”

The real proof, he concludes, is in quiet success: releases ship on time, stakeholders stop asking which number is true, and even users on slow connections still get the right data fast.