Mon – Sat: 9:00 AM – 7:00 PM IST
Implementation

ERP implementation phases

What actually happens between signing and go-live, what each phase should produce, and where the time goes.

In short

An ERP implementation runs through discovery, design, configuration and build, data migration, testing, training, cutover and hypercare. The phases overlap in practice, but each produces something specific. Projects slip when a phase is declared complete without its deliverable existing — most often design sign-off without a documented design.

Why It Matters

Why this matters in practice

The phases are not the interesting part; the exit criteria are. A project that moves from design to build because the calendar says so, rather than because the design is agreed and written down, will rediscover the design during build at several times the cost. Almost every overrun traces back to a phase that was exited early.

Each phase has a deliverable, not just a date
Data migration starts early, not at the end
Testing uses real data and real users, not sample records
Hypercare is part of the plan, not goodwill afterwards
How It Works

How it works in practice

Discovery

Current processes, systems, data and pain points documented. Produces a scope and a fit-gap, not a wish list.

Design

How the system will be configured, what will be customised and what the business will change. Signed off in writing before build starts.

Build and configure

Configuration, extensions, integrations and reports, delivered in increments the business can see.

Data migration

Extract, cleanse, map, load, reconcile — repeatedly. Begins in the first weeks and runs in parallel throughout.

Test, train, cut over

Integration and user acceptance testing against real data, training on the configured system, then a rehearsed cutover and hypercare.

Worked example

A worked example: where a six-month project spends its time

A mid-market ERP implementation budgeted at six months breaks down, in practice, quite differently from how it is usually planned.

Discovery and design take around six weeks and are almost entirely decision-making rather than software. Configuration and development take around ten weeks and are the part everyone pictures when they imagine the project. Data migration takes around eight weeks running in parallel, and it is the activity that most often slips. Testing takes around six weeks including at least two full cycles. Training and cutover preparation take around four. Hypercare runs four to six weeks after go-live.

The two observations that matter: migration and testing together exceed configuration, and both depend heavily on the client's own effort rather than the implementer's. Plans that allocate most of the calendar to configuration and treat migration as a late task are the ones that overrun. If you want to know whether a plan is realistic, look at what it assumes about data cleansing and who is doing it.

Key terms

The vocabulary, briefly

Term What it means
Discovery Current-state mapping and fit-gap, producing scope rather than a sales proposal.
Conference room pilot Walking real business scenarios through the configured system before build completes.
Cutover The transition from old system to new, rehearsed before it is performed.
Hypercare Intensive support immediately after go-live, with the build team still engaged.
Benefits realisation Checking afterwards whether the business case actually happened. Routinely skipped.
What Goes Wrong

Common mistakes worth avoiding

Design signed off verbally

If it is not written down, it will be remembered differently by each party during build.

Data migration left to the end

The most common single cause of a delayed go-live. Cleansing is the client's work and takes months.

Testing with sample data

Sample data passes. Real data contains the twenty years of oddities that break things.

No hypercare period

Go-live is when problems surface. A team that disbands on the day leaves them with nobody.

What to measure

Numbers worth tracking

Metric What it tells you What distorts it
Open design decisions Decisions waiting on the business. The most common cause of schedule slip.
Test cases passed on first execution Build quality. Low rates early mean the design was not understood.
Migration rehearsal count Two or three full rehearsals before cutover is normal; fewer is optimistic.
Defects found after go-live Testing effectiveness, and the thing hypercare is sized against.
FAQ

Questions we are asked about this

It varies with scope, data quality and how much the business is prepared to change. The honest answer requires a discovery phase — anyone quoting a duration before seeing your data is guessing.
Data migration and scope change, in that order. Both are manageable with early data work and a written design that change requests are measured against.
They must, or the elapsed time becomes unreasonable. Data migration runs alongside configuration; training material is written during testing; cutover planning starts long before cutover. What should not overlap is building against a design that is not yet agreed — that produces rework at a rate that exceeds the time saved. The discipline is to overlap activities with different dependencies, not to start everything at once.
Walking genuine business scenarios end to end through the configured system with the people who actually do the work, before build is complete. It is the cheapest point at which to discover that a process does not fit, because changing configuration is still easy and nothing has been migrated. Projects that skip it find the same problems during user acceptance testing, when they are substantially more expensive.
Through at least one full business cycle, which for most organisations means one complete month-end close plus any periodic process such as payroll or statutory filing. Four to six weeks is typical. Ending hypercare before the first close is a false economy, because the close is where the problems that matter surface and where the people who built the system are the only ones who can diagnose them quickly.
Data migration, consistently and by a wide margin. The work is not the loading; it is deciding what to bring, cleansing it, mapping it to a different structure, reconciling it and rehearsing the whole sequence. Much of that work can only be done by people who know the data, which means your staff rather than the implementer's. Any plan that does not name who is doing the cleansing has not confronted the largest risk in the project.
Related

Where to read next

ERP data migration

Read more

What is UAT?

Read more

Our ERP software solutions

Read more

Need this applied to your own systems?

We build and support the systems this page describes. Start with a free working proof of concept — no cost, no commitment.