The corporate data platform

The hardest decision in a data platform is rarely technical. It is to stop treating each source, survey or report as its own product. Twenty-five surveys from several institutes, each enriched separately, each with its own questionnaire, cannot be compared; one common zone that ingests once, enriches once and maps questions and scores to shared tags can — at the price of a slower first delivery and one more layer between the source and the business.

That price must be accepted explicitly, and the mapping across heterogeneous sources made the success condition, proved first on two pairs before being generalised. Governance follows the same logic: single IT responsibility, shared business governance, a stable exchange format with external providers as a prerequisite.

A platform is also a reconciliation engine. When the official revenue report cannot be audited and conflicts with transaction-level dashboards, every margin discussion becomes a dispute over whose numbers are credible. Rebuilding volume, fees and interchange from transaction grain, defining seventeen dimensions with owners and lifecycles, and leaving a threshold-based reconciliation control that Finance can run alone replaces debate with evidence — within 0.04% on volume, in one acquirer’s case.

Platforms are killed by budget cuts more often than by technology. What survives a cut is what was argued on continuity, compliance and visible cost rather than on a return-on-investment case: the unified model, the central translation, the quality indicators that stop being computed if the platform stops. A structured retreat is a legitimate outcome; design for it.