DataEdge is the control room for an ERP and HCM data conversion. It governs every step from discovery to cutover, and proves afterwards that what arrived in Oracle Fusion is what was meant to arrive.
An ERP conversion is a governance exercise run under time pressure, in which dozens of people make thousands of decisions. Programmes fail their audits and miss their cutover windows for the same three reasons every time.
Spreadsheets and scripts
With DataEdge
The discipline
Anything DataEdge presents as a measurement is computed when it is displayed. Only decisions are persisted. A status cannot be stale because there is no stored status, and a readiness percentage cannot be optimistic because nobody can type it in.
Approvals, scope calls, waivers and sign-offs are the only things DataEdge persists, each with the person, the time and the evidence.
Delete a mapping and every figure that depended on it disappears on the next view: its dataset, validation results, load package and readiness. A platform that kept showing them would be lying.
An entity no load has reached reports “no load run has reached this entity”, not zero. Post-load checks report “not run”, never “pass”.
Architecture
CommandEdge owns connectivity, credentials, data movement, storage and masking. DataEdge is the control room: it decides what runs, in what order and against which environment, and shows the result.
The systems of record you are leaving, such as PeopleSoft HCM and Financials. Read only, through CommandEdge. Never written to.
One governed store, organised by business entity. It holds two complete record sets in one canonical shape: the legacy set, folded from every contributing source, and the new-system set as built for production. The difference between them is the conversion, stated as a number.
HCM, Payroll, Time & Absence and ERP, production and test pods. Written through Oracle’s loaders, and read back after go-live, when production becomes a source in its own right.
Rigid in, flexible out. The inbound leg lands every build in the same canonical columns, so one release can be compared with the next. The outbound leg is generated per destination, and every outbound rule carries its inverse, so read-back fidelity is a measured property of the design rather than an assurance.
Twenty modules, six clusters
The clusters are a delivery sequence as well as a grouping, and the dependency is real: no repository dataset exists without a mapping, no load package without a validation result, and no reconciliation without a load run.
Discovery & assessment, data profiling and data governance. The system inventory, scope decisions with their rationale, legacy control totals, and every column measured rather than trusted.
The MirrorsEdge central repository, the mapping repository and versioned crosswalks. Transforms resolve against the approved version in force, never the working copy.
The cleansing framework, the conversion rule engine and the load package builder. Rules tied to the anomalies they answer, inverses derived, packages sequenced into dependency waves.
The validation framework, load execution and error management. Pre-load, outbound and post-load checks; validate-only, load and retry modes; a finding register generated from the evidence.
Reconciliation, testing, environment management and production cutover. A generated runbook, a computed critical path, and a go/no-go verdict evaluated against live state.
Audit & compliance, the security model, analytics & reporting, the knowledge repository and AI-assisted proposals, all assembled from live state.
Prove & control
Everything before reconciliation moves data. This is where DataEdge proves it arrived, that the business can work with it, and that the programme is ready for the night.
Every legacy record is accounted for as accepted, rejected, never submitted, held on exception or unexplained. Sign-off is gated on the arithmetic: where an entity cannot be signed, the reasons are printed in place of the button.
Nine criteria, each a function over live state, each printing its current value and a route to the screen that would change it. A sponsor may approve against an unmet criterion, and the criteria they overrode are recorded by name.
Eleven audit deliverables, from mapping specifications to the final go-live approval, each generated from live state on demand. Approval is blocked while any section carries no evidence, and the refusal names the section.
The verdict is not a status somebody sets before a steering meeting. It is what the evidence says, and it will say go when the evidence does.
Governance built in
Writing a mapping, drafting a rule or proposing a cleansing step needs no authority, because each is reviewed at a gate. What is gated is the act that changes what the programme is committed to.
Clear boundaries
Separation of concerns is how we build. DataEdge does one job, and it is honest about the edges of it.
It does not own connectivity, credentials or data movement. Those belong to CommandEdge. DataEdge decides what runs and shows the result.
It profiles and cleanses in service of a conversion, against a known target shape, not as a business-as-usual monitor.
Its analytics exist so value arrives early and conversion questions are answered with data, not to replace Oracle Fusion’s own reporting after go-live.
Every inference is a proposal addressed to a named owner. People decide, and DataEdge records who did.
Tell us about your programme: the systems you are leaving, the Fusion pillars you are moving to, and your cutover window. We will show you how DataEdge would run it.
Built for Oracle Fusion HCM, Payroll, Time & Absence and ERP · orchestrated by CommandEdge
