ERP separation is the deepest migration in any carve-out.

The three ERP separation approaches.

There are three credible approaches. Clone the seller's ERP into a separate Newco instance. Implement a fresh ERP from scratch. Adopt a SaaS ERP designed for mid-market companies. Each has different cost, risk, and timeline. Each is appropriate for a specific situation, and the decision is made in the first 60 days after close.

Cloning the seller's ERP is the fastest path. The Newco gets a copy of the seller's configuration, master data, and historical transactions. Stand up time is 8 to 12 months. The cost is moderate. The downside is that the Newco inherits the seller's configuration, which may not fit the Newco's operating model. Cloning is the right answer when the Newco is operationally similar to the parent and the priority is exit speed.

Fresh implementation is the longest path and the most expensive. The Newco designs its own ERP configuration, migrates only the data it needs, and exits the TSA at a clean operating point. Implementation runs 18 to 24 months. The cost can be three to five times the cloning approach. Fresh implementation is the right answer when the Newco's operating model differs materially from the seller's, or when the seller's ERP is being sunsetted.

SaaS adoption sits between the two. A modern cloud ERP can be implemented in 9 to 14 months for a mid-market Newco. The configuration is constrained by the platform, which is often a feature rather than a limitation. SaaS adoption fits Newcos that want to skip a generation of technology and that can accept the platform's standard processes. The decision drives every downstream date.

Data conversion is the single biggest risk.

Data conversion is the work that determines whether the new ERP works on Day One. Master data conversion includes customer, vendor, product, item, employee, asset, account. Open transaction conversion includes open purchase orders, open sales orders, open invoices, open receipts, open work orders. Historical balance conversion includes opening trial balance, opening subledger balances, and reference history.

The data work is heavier than the configuration work. A typical ERP separation has 40 to 60 percent of the budget in data conversion. The volume scales with entity count and master data depth, not with implementation approach. Skimping on data conversion is the single most common reason for a slow Newco close in the first three months after cutover.

Data quality in the source system is usually worse than the seller's IT team represents. Duplicate vendor records, inactive customers carried forward, items with broken bill of material references, asset records missing depreciation history. Each issue requires a decision before conversion. The decisions accumulate into a working data conversion plan that runs four to six months and consumes a meaningful share of the program budget.

Mock conversions are the discipline that keeps data work on track. Run a full conversion, validate every table, document every variance, fix the conversion logic, run again. Three to five mock conversions are the right rhythm. By the final mock, every variance should be either explained or fixed. A botched first conversion at cutover is the alternative.

The integration footprint is rarely documented.

Every ERP has more integrations than the documented list suggests. Outbound feeds to data warehouses, inbound feeds from time and attendance, file based connections to payroll, real time interfaces with EDI vendors, batch interfaces with banks for payment files. A typical ERP has 30 to 80 integrations. The seller's documentation usually captures 60 to 75 percent of them.

The buyer's first deliverable in the ERP workstream is a complete integration inventory. Every interface, every endpoint, every protocol, every business owner, every cutover plan. Without a complete inventory, the cutover finds gaps in production, which is the worst place to find them. The inventory work runs 60 to 90 days into the program.

Integration rebuild is rarely a copy operation. The endpoints change. The authentication changes. The data formats sometimes change. Some integrations need a full redesign because they were built against legacy assumptions that no longer apply. Building rebuild time into the program plan for each integration is the only way the cutover lands on schedule. The full milestone framework is in TSA exit milestones, explained.

Third-party vendor coordination is a hidden integration cost. Banks, EDI partners, customers, freight providers, all have to be coordinated. Each external party has its own lead time for testing and cutover. The buyer's program manager has to maintain the external coordination plan alongside the internal cutover plan or external partners become a Day One issue.

Cutover weekend is a fixed event.

ERP cutover is rarely incremental. A big bang cutover, where the seller's ERP stops accepting transactions on Friday evening and the Newco ERP starts on Monday morning, is the dominant approach for a reason. Incremental cutovers create reconciliation complexity across the weekend window and rarely deliver the speed advantage they appear to promise.

The cutover weekend itself runs to a minute level plan. Final data extraction Friday evening. Master and open transaction load Saturday morning. Reconciliation Saturday afternoon. Integration testing Sunday morning. Final go or no go decision Sunday afternoon. First business day operations Monday morning. The plan is rehearsed twice in advance with no shortcuts.

The no go decision is the most under exercised muscle in ERP cutover. Most programs assume the cutover will succeed. The right discipline is to define the no go criteria in writing, three weeks before the cutover. If conversion reconciliation shows a variance above a defined threshold, if integration testing fails on a critical interface, the program leadership reverts to the rollback plan rather than pushing through. Rollback plans exist for a reason.

Hypercare follows cutover. The two weeks after cutover require dedicated support across every workstream. Finance close runs longer than steady state. Order to cash transactions get held up at any configuration gap. Daily standups, named owners by issue type, and direct access to vendor support are all standard. The hypercare team should be sized at 150 percent of steady state for the first two weeks and 120 percent for the second two weeks.

Cost ranges and where overruns hide.

ERP separation costs vary widely with approach. A clone implementation for a mid-market Newco runs $3M to $8M in implementation cost plus TSA fees during the cutover period. A SaaS adoption runs $5M to $15M. A fresh ERP implementation runs $10M to $30M and can go higher for global operations. The seller's TSA fees for ERP hosting alone range from $40K to $250K per month, separate from the implementation budget.

Overrun pressure concentrates in three places. Data conversion, where source data is worse than expected. Integration rebuild, where undocumented interfaces surface late. Hypercare, where the support load is heavier than planned and runs longer than the two week window. Each of these adds 10 to 30 percent to the headline budget if the program is not disciplined about scope and rollback. Workstream by workstream cost ranges are in TSA exit cost benchmarks.

The systems integrator decision is a budget decision as much as a quality decision. The right integrator for an ERP separation has done the same approach with the same platform multiple times before. Cheaper integrators learn on the buyer's dollar. More expensive integrators with the right experience often finish faster and cost less in total. The decision is made on track record, not on day rate.

Date discipline is the saving. Holding the cutover date and adjusting scope is what preserves the budget. A six month delay on an ERP cutover costs $2M to $5M in extension fees plus operational drag. The cost of cutting wave four scope to protect the wave three core is almost always lower than the cost of slipping the entire program.

More on TSA exit strategy.