ERP Implementation Thought Leadership

Why the platform choice belongs to the CEO, not the CIO alone

An ERP replacement typically consumes 1.5 to 3 percent of annual revenue and reshapes how finance, supply chain, and operations record every transaction for the next decade. For a company at 400 million in revenue, that is a 6 to 12 million dollar program before the first efficiency arrives. That makes it a strategy decision wearing a technology costume. When boards abdicate the choice to a procurement scoring sheet, they inherit a system optimized for the lowest evaluation friction rather than the business they actually run.

The two dominant options resolve to a philosophy question, not a feature comparison. SAP S/4HANA rewards organizations with deep process complexity, heavy manufacturing, multi-plant costing, and a tolerance for configuration depth. Oracle Fusion Cloud rewards organizations that want a faster, more standardized cadence, with quarterly updates absorbed centrally and less appetite for bespoke process. The wrong fit is rarely fatal, but it routinely costs 12 to 18 months of avoidable rework, a blown-out change budget, and a demoralized finance team that stops trusting the numbers. The decision that matters is made in the first two weeks, when the board decides whether it is buying a configurable platform to fit its process or adopting a standard process and reshaping the business around it.

A 180-day phase plan with exit gates you can defend

Treat the program as five phases, each ending in a gate the steering committee must sign before funding the next. No gate, no spend. This converts a vague multi-quarter effort into a set of falsifiable checkpoints that a non-technical director can hold the team to. The point is not to hit the calendar, it is to refuse to advance on an unsound foundation.

The discipline is in the gates, not the calendar. If the second mock load leaves a 40,000 dollar unexplained variance on a control account, you do not proceed to cutover rehearsal on schedule, no matter how much the plan says you should. Slipping a gate by three weeks is far cheaper than reversing a bad go-live, which can strand a full close cycle and force a manual parallel run.

What the steering committee should mandate

• Set a standard-first rule: every requested customization must be defended before the committee, with the default answer being no. Aim to keep custom objects under 20, because each RICEF item adds an estimated 8,000 to 15,000 dollars of lifetime maintenance.
• Appoint a single accountable business owner per stream, finance, supply chain, order-to-cash, not just an IT project manager, and give each of them sign-off authority on their own gate.
• Fund data migration as its own workstream with its own lead and budget, because migration is where roughly 40 percent of go-live failures originate and it is the most under-resourced phase.
• Define hypercare SLAs before go-live: critical incident response under 30 minutes, a staffed war room for the first 15 business days, and an explicit demobilization trigger tied to ticket trend, not to a date.
• Decide the composable question early: keep the ERP as a clean digital core and decouple fast-changing capabilities like pricing, tax, and CPQ into best-of-breed services behind clean APIs, so the next change does not require a core upgrade.

Where ERP programs quietly go wrong

• Migrating history you will never use. Fix: agree a data retention rule, typically two years of open items plus balances, and archive the rest outside the ERP where it is cheap to hold and easy to query.
• Letting the system integrator own the timeline. Fix: the client chairs every gate review and holds the go or no-go decision, so the incentive to declare victory sits with the party that lives with the result.
• Customizing to preserve legacy quirks. Fix: change the process to match the standard unless a customization protects a genuine competitive edge or a hard legal obligation, and require a written justification for each exception.
• Treating training as a launch-week event. Fix: run role-based training two weeks before cutover using real transactions in a sandbox, then reinforce daily during hypercare while support is still on the floor.
• No clear ownership after go-live. Fix: stand up a product-owner model with a funded run team, so the ERP evolves through a backlog rather than ossifying into the next legacy system.

Five checks before you approve the business case

• Confirm the platform fit maps to your business shape, not the sharpest demo you were shown.
• Verify the plan has five gates with a named, accountable business owner on each.
• Cap the customization backlog in writing at 20 items and enforce it at every review.
• Require two clean mock data loads, reconciled to the penny, before authorizing cutover.
• Lock hypercare SLAs and a trend-based demobilization trigger before go-live day.

Related advisory guides

Go deeper on this theme with the detailed guides in this collection.