ERP Implementation Guide: Phases, Roles and Risks
By Vikas Saroj, ERP, Digital Transformation & Growth Consultant
Key takeaways
- A successful ERP implementation follows eight phases: discovery and planning, solution design, configuration and build, data migration, testing, training, go-live and cutover, and optimization.
- Before starting, settle why you are implementing, what is in scope for the first release, and which business owner with real authority owns the project.
- For each requirement gap, prefer adopting the standard process, then configuration, then integration, and treat customization as the last choice because it must be maintained through upgrades.
- User acceptance testing should be run by the business with real end-to-end scenarios and migrated data, against exit criteria agreed in advance.
- For most SMB and mid-market businesses, a phased go-live with finance and core operations in the first release is safer than a big-bang launch.

A successful ERP implementation follows a clear sequence: discovery and planning, solution design, configuration and build, data migration, testing, training, go-live, and post-launch optimization. Each phase has a defined output and a decision point before the next begins. Most ERP projects that struggle do so because of unclear scope, poor data or weak ownership, not because the software fails.
This guide walks through the phases I use with clients, the roles you need, and the risks to watch. It applies whether you are implementing Zoho, Odoo, ERPNext or a larger suite.
Before you start: three things to settle
- Why are we doing this? Write down the business problems the ERP must solve, such as a slow month-end close, unreliable stock data or no visibility of project margins. These become your success criteria.
- What is in scope for the first release? Decide which entities, modules, locations and processes go live first. Everything else goes in a later phase.
- Who owns it? An ERP project needs a business owner with authority, not just an IT lead.
If you have not yet selected a platform, start with how to choose an ERP.
The ERP implementation phases
| Phase | Main activities | Key output |
|---|---|---|
| 1. Discovery & planning | Process workshops, requirements, scope, plan, governance | Signed-off scope, BRD and project plan |
| 2. Solution design | Fit-gap analysis, future-state processes, data model, integration design | Solution design document |
| 3. Configuration & build | System setup, workflows, approvals, reports, integrations, agreed customizations | Configured system in a test environment |
| 4. Data migration | Data cleansing, mapping, trial loads, reconciliation | Validated migration scripts and templates |
| 5. Testing | Unit, integration and user acceptance testing | UAT sign-off and resolved defects |
| 6. Training & change | Role-based training, user guides, communication | Trained users and go-live readiness |
| 7. Go-live & cutover | Final data load, opening balances, switch-over, hypercare | Live system, stable operations |
| 8. Optimize | Fix issues, refine reports, add automation, plan next phase | Improvement backlog and phase-two roadmap |
1. Discovery and planning
Run workshops for each major process: order-to-cash, procure-to-pay, inventory, production or projects, and record-to-report. Capture how work happens today, where it breaks, and what needs to change. Document requirements in a business requirements document; see how to create an ERP BRD.
Agree the project plan, milestones, governance (who decides what, and how fast), and a change-request process. Scope creep starts here if this step is rushed.
2. Solution design
Compare requirements against the platform's standard capabilities. For each gap, choose one of four options: adopt the standard process, configure, integrate with another tool, or customize. Customization should be the last choice, because every custom feature must be maintained and tested through upgrades.
Design the future-state processes, chart of accounts, item and customer master structure, user roles and permissions, and integration points. Get business sign-off before building.
3. Configuration and build
Set up the company structure, taxes, currencies, warehouses, workflows, approval rules, document templates and reports. Build integrations to systems such as CRM, e-commerce, banking or payroll. My system integration work typically happens in this phase.
Work in short cycles and show progress to process owners every week or two. Waiting until the end to show users the system is how surprises happen.
4. Data migration
Data migration is consistently underestimated. Decide what to migrate (master data, open transactions, opening balances, and how much history), then cleanse it at the source. Remove duplicates, fix inconsistent units and codes, and archive inactive records.
Run at least two trial loads and reconcile each one against the old system. The ERP migration checklist covers this in detail.
5. Testing
Testing runs in layers:
- Unit testing: each configuration and customization works on its own.
- Integration testing: data flows correctly between modules and connected systems.
- User acceptance testing (UAT): real users run real end-to-end scenarios with migrated data and confirm the system supports their work.
UAT should be done by the business, not the implementer. Agree exit criteria in advance, for example: no open critical defects and all key scenarios passed.
6. Training and change management
Train by role, not by module. A warehouse picker needs to know three screens well, not the whole inventory module. Use the migrated test data so training feels real. Identify "super users" in each department who can support colleagues after go-live.
Communicate early and often about what is changing, why, and when. Resistance usually comes from uncertainty, not from the software.
7. Go-live and cutover
Plan cutover as a detailed checklist with owners and timings: freeze transactions in the old system, load final balances and open items, reconcile, then open the new system. Choose a go-live date that avoids peak trading and, ideally, aligns with a period start.
Plan for a hypercare period immediately after go-live, with the implementation team on hand to fix issues quickly. Expect questions and minor problems; that is normal.
8. Optimize and scale
Once operations are stable, review what is working, close remaining gaps, add automation where manual steps remain, and plan the next phase of modules or locations.
Roles and responsibilities
| Role | Responsibility |
|---|---|
| Executive sponsor | Owns the business case, removes blockers, makes final decisions on scope and priority |
| Project manager (client side) | Coordinates internal resources, tracks plan and risks, manages decisions and sign-offs |
| Process owners | Define requirements for their area, approve design, lead UAT for their processes |
| Super users | Test, help train colleagues, become first-line support after go-live |
| Implementation partner or consultant | Designs, configures and builds the solution, advises on best practice, supports migration and go-live |
| IT / systems lead | Infrastructure, security, access, integrations and technical support |
| Data owner | Responsible for data quality, cleansing and migration sign-off |
In smaller businesses one person may hold several roles, which is fine as long as each responsibility is explicitly assigned and that person has time freed up to do it.
Common risks and how to reduce them
- Scope creep. New requests keep arriving mid-project. Use a change-request process and park non-essentials for phase two.
- Over-customization. Recreating old habits in the new system adds cost and upgrade risk. Challenge every customization request with "what business problem does this solve?"
- Poor data quality. Bad data loaded into a new system produces bad reports faster. Start cleansing early.
- Key people not available. Process owners are busy running the business. Backfill their routine work or the project will stall.
- Weak testing. Skipping or rushing UAT moves defects into production. Protect testing time in the plan.
- Low adoption. If people keep using spreadsheets after go-live, the ERP becomes an expensive second system. Make the ERP the only official source for key data and track usage.
- Big-bang overload. Going live with everything at once raises risk. Phasing by module, entity or location is often safer.
Phased vs big-bang go-live
A big-bang go-live switches everything on at once. It avoids temporary integrations between old and new systems but concentrates risk. A phased approach rolls out by module, entity or site, which spreads risk and lets the team learn, but may need interim workarounds. For most SMB and mid-market businesses I recommend phasing, with finance and core operations in the first release.
Next steps
If you are planning an ERP implementation, or one is underway and not going to plan, I can help with scoping, design, data migration and go-live readiness. Read more about my ERP consulting approach or book a consultation.
Frequently Asked Questions
What are the main phases of an ERP implementation?
The typical phases are discovery and planning, solution design, configuration and build, data migration, testing, training and change management, go-live and cutover, and post-launch optimization. Each phase should end with a clear output and sign-off before the next begins.
How long does an ERP implementation take?
It depends on scope: the number of modules, entities, locations, integrations, customizations and the quality of your data. A focused first phase for a small business is much shorter than a multi-country rollout. A proper discovery phase is the only reliable way to estimate your timeline.
Who should own an ERP project?
The business should own it, with an executive sponsor who has authority over scope and priorities. IT and the implementation partner play essential roles, but process owners from finance and operations must define requirements, approve the design and lead user acceptance testing.
Is a phased or big-bang go-live better?
For most small and mid-sized businesses, a phased go-live is safer because it spreads risk and lets the team learn between phases. Big-bang can work when processes are simple, data is clean and interim integrations would be costly.