Middleware is the plumbing. Rebuild it before the cutover.

Why the middleware layer is the seam. Every system talks through it.

A modern enterprise integration platform sits between the operational systems and everything that consumes data from them. Orders flow from the order management system into the ERP, then into the warehouse, then into the billing platform, then into the CRM, then into the analytics layer. Each hop is a contract maintained by the integration layer. In a carve-out, every one of those contracts has to survive the seam.

The seller's middleware was built to serve the whole enterprise. It holds connectors to systems Newco will never see again, business rules that mix Newco and seller logic, and routing policies that assume the old org chart. None of it transfers cleanly. The buyer-side advisor pressure tests the inherited integration estate against the standalone Newco perimeter in month one and builds a realistic rebuild roadmap before TSA pricing is finalized.

The right framing is that the integration layer is the seam at the orchestration level. The seam needs deliberate engineering, not a copy of the seller's runbook. The work pairs closely with carve-out data warehouse separation because most pipelines run through the same platform.

Inventory of every integration in scope. If it moves data, it gets cataloged.

The first deliverable is a complete integration inventory. Every API call, every batch file, every event stream, every scheduled job that touches a Newco system. The inventory pulls from the seller's middleware tool, from application logs, from network traffic, and from interviews with the people who run the integrations. The buyer-side advisor leads the inventory because the seller's TSA documentation usually undersells the volume of point to point integrations sitting outside the official platform.

Each integration carries a small set of attributes that drive the rebuild decision. The source system, the target system, the data domain, the frequency, the volume, the business owner, the criticality, and whether the integration is shared with the seller or specific to Newco. The catalog becomes the single source of truth for sequencing and the basis for the integration service catalog in the TSA itself.

Once the catalog is complete, the buyer-side advisor segments the integrations into four buckets: rebuild on Newco platform, retire because the underlying process is changing, replace with a SaaS native connector, or defer until after exit. Roughly half of inherited integrations rationalize in one of the latter three buckets, which is the first cost takeout opportunity in the workstream.

Picking the Newco platform. Match the architecture to the standalone scale.

The platform decision is shaped by Newco's standalone scale, the dominant application stack, and the team that will run it. The mainstream options are MuleSoft, Boomi, Workato, Informatica, Azure Logic Apps, AWS EventBridge plus Step Functions, and increasingly Snowflake or Databricks native pipelines for analytical flows. The seller's choice was right for the seller's scale. Newco usually needs something lighter and faster to operate.

A common decision split is operational integrations on one platform and analytical integrations on another. Operational flows that need synchronous APIs, transactional semantics, and tight latency live on the integration platform proper. Analytical flows that move batches into the warehouse live on an ELT tool such as Fivetran, Airbyte, or the warehouse native loaders. Mixing both on one platform usually leaves Newco overpaying for capability it does not use.

The buyer-side advisor brings the operating partner into the platform selection because the choice has implications well beyond the carve-out. Newco will live with this platform through the value creation plan and into the next transaction. The work pairs with carve-out application portfolio rationalization so the integration choice fits the broader application strategy.

Sequencing the rebuild against TSA exit. Critical paths first, everything else later.

Sequencing is the difference between a rebuild that finishes on time and one that drifts into extension fees. The integrations that gate other workstreams have to move first. The order to cash integration gates the finance separation. The HR feed to payroll gates the payroll cutover. The warehouse load gates the analytics rebuild. The buyer-side advisor maps these dependencies and freezes the integration sequence against the master exit plan in month three.

Below the critical path are the integrations that can move on a slower track. Marketing automation feeds, contract management updates, reference data syncs. These get a longer runway and tighter scope. The temptation to rebuild every integration in parallel is the most expensive mistake at the platform level. Engineering capacity is finite and quality drops when too many flows are in flight at once.

The sequence is also reflected in the TSA itself. The integrations the seller continues to provide are tied to a service catalog entry, a price, and an exit milestone. As each integration moves to the Newco platform, the corresponding TSA service is sunset and the cost drops out of the run rate. The work ties directly to TSA exit milestones.

Testing, parallel running, and cutover. No integration moves without reconciliation.

An integration cutover that goes wrong corrupts data on both sides of the seam. The reconciliation framework has to be in place before any production flow moves. The buyer-side advisor designs a parallel running window for each critical integration where the seller flow and the Newco flow both run against the same source, with daily diff reports against agreed tolerances. Variances are investigated, root caused, and either the Newco flow is patched or the variance is documented as an accepted difference.

Cutover itself happens at the end of a parallel period where the reconciliation has been clean for several consecutive cycles. The seller flow is disabled, the Newco flow becomes primary, and a defined rollback window runs in case downstream consumers report anomalies. Each cutover has a named owner, a defined backout plan, and a documented signoff from the receiving business unit.

When all critical integrations have cut over and reconciled clean, the seller's middleware can be retired from Newco's perspective and the related TSA service exits. Both parties countersign the integration separation as complete, which is a documented exit milestone in the TSA. The work pairs with TSA IT separation.

Operating the platform after exit. The team carries the runbook forward.

Newco needs more than a running platform. Newco needs a team that can operate it without the seller's support desk. The buyer-side advisor designs the operating model in parallel with the build, defining the on call rotation, the change management process, the monitoring and alerting standards, and the documentation library. The handover is graduated. The build team operates the platform for the first weeks after exit. The standing Newco team takes over once the operating runbook has been exercised.

Documentation is the artifact that survives the team turnover that always follows a carve-out. Every integration gets a one page runbook covering the source, target, schedule, owner, escalation contact, and known failure modes. The runbooks are stored where the Newco operations team works, not buried in a knowledge base the team will not find. The buyer-side advisor enforces the documentation standard as a precondition for any integration to be considered production complete.

The operating model carries forward into the value creation plan. The integration platform is one of the highest leverage assets in the Newco technology estate because every new initiative depends on it. A well operated platform accelerates the next ERP rollout, the next data product, and the next acquisition. The work pairs with operating partner TSA vendor management.

More on carve-out technology.