How to Plan an ERP Implementation
The warehouse lead says the system shows 40 units on hand. The shelf says 31. Sales promised 35 to a customer an hour ago. Finance will close the month in nine days and the inventory value on the books is a number nobody quite trusts. This scene is why companies buy an ERP, and it is also why so many ERP projects go badly. The system does not fix the disagreement. It only makes the disagreement visible, faster, to more people.
ERP stands for enterprise resource planning. In plain terms it is one system of record for the money and the stuff, covering general ledger, purchasing, inventory, production or fulfillment, sales orders, and invoicing. If you run a manufacturer or a distributor with 20 to 200 people, an ERP implementation is likely the largest operational project you will undertake in a decade. This guide covers how to plan it so that the first month-end close after go-live is boring.
Decide what the ERP is for before you decide what it does
Start with a short statement of purpose that the whole leadership team can repeat. Good ones are specific. "Know true inventory by location every day, ship orders without re-keying, and close the books in six working days." Poor ones are vague, such as "modernize operations."
From that statement, derive three to five measurable outcomes. For example.
- Inventory accuracy at cycle count above an agreed percentage.
- Order entry without retyping between quoting, orders and invoicing.
- Month-end close in a stated number of working days.
- One customer price list in one place.
You will use these outcomes later to settle arguments. When someone asks for a customization, the question is which outcome it serves.
Scope is a set of decisions, not a feature list
Scope has four dimensions, and the project goes wrong when only one is discussed.
- Functional scope. Which processes go into the ERP at go-live, and which stay outside for now. Finance, purchasing, inventory and sales orders are a typical first wave. Advanced planning, quality management or a service module can come in a second phase.
- Organizational scope. Which entities, plants, warehouses and currencies. Each additional legal entity multiplies chart-of-accounts, tax and intercompany work.
- Data scope. How much history to bring. Opening balances and open transactions are essential. Ten years of closed invoices usually are not, and can be archived separately and made searchable.
- Integration scope. Which other systems must connect, such as e-commerce, a shipping tool, payroll, a CRM, EDI with large customers, or a manufacturing floor system.
Write scope as in and out lists, and keep an explicit "later" list. A "later" list is a pressure valve. People accept a deferral far more easily than a flat no.
Phase, but not too thin
Phasing reduces risk, yet every phase has a cutover. Two phases are common. Three is a lot. A very common mistake is to go live on finance alone and defer inventory, which leaves you with a ledger that cannot see the largest asset on your balance sheet.
Master data is the project
Master data is the slow-changing reference information that every transaction points to. Items or parts, bills of materials, customers, vendors, locations, units of measure, price lists, the chart of accounts. Transactions are the daily events, such as a purchase order or a shipment.
ERP projects that miss their dates nearly always miss because of master data, not software. The system will faithfully multiply any inconsistency you load.
Give every master record an owner
For each data domain, name one person who decides what is correct. The purchasing manager owns vendors. The operations lead owns items and bills of materials. Finance owns the chart of accounts and customers' credit terms. Without owners, cleanup becomes a committee, and committees do not delete things.
Clean for the keys that matter
Some typical problems and fixes.
- Duplicate parts with different codes. Decide on an item numbering scheme and merge.
- Units of measure that disagree, for example purchased by the case and sold by the each. Define conversion factors and test them.
- Bills of materials that reflect how things were built three years ago. Walk the floor and confirm.
- Customers with several spellings and addresses. Choose a matching key, such as tax ID or normalized name plus address, and a survivorship rule for conflicts.
- Obsolete vendors and items. Flag as inactive and do not load.
A small worked example, illustrative numbers
Here is a hypothetical. A distributor has 6,500 item records in its legacy system. During cleanup it finds roughly 900 that have had no movement in three years, about 400 duplicates caused by supplier code changes, and 250 with missing units of measure. Loading 6,500 messy records would carry the problems forward. Loading about 4,950 verified records gives the new system a clean start, and the 1,550 excluded rows are stored in an archive for lookup. Those numbers are made up for illustration, but the ratios, with a fifth to a third of item records needing attention, are plausible for a company that has never pruned its catalog.
Fit-gap analysis without drama
A fit-gap analysis compares what the ERP does out of the box against what your processes need. The output is a list of requirements, each marked as a fit (standard function does it), a configuration (a setting or workflow does it), a gap (needs customization or an add-on), or a change (you alter your process to match the software).
Run it from your real scenarios
Do not run fit-gap from the vendor's feature brochure. Write 10 to 15 scenarios from your own operation and walk them through the demo environment with your own data. Useful scenarios for a manufacturer or distributor include the following.
- A customer orders a configured product with a special price and a split shipment.
- A supplier ships short and you must receive partially and keep the purchase order open.
- A returned item goes back into stock at a different cost.
- A work order consumes materials with a scrap percentage.
- A price change takes effect on a future date.
- A landed cost, such as freight or duty, is allocated back to inventory value.
Default to the standard
Every customization is a cost you pay at build, at testing, at each upgrade and in each training session. Create a rule that a gap needs an executive sponsor's signature and a stated business outcome it supports. For many gaps, changing the process is cheaper and better. Keep customizations for the places where your process is truly a competitive advantage, which is a short list.
| Requirement type | Typical action | Who decides |
|---|---|---|
| Fit | Use standard, document | Process owner |
| Configuration | Set up, test, document | Process owner and implementer |
| Process change | Redesign, train | Executive sponsor |
| Gap, customization | Cost it, justify it, approve it | Executive sponsor |
| Gap, third-party add-on | Check vendor support and upgrade path | Project lead |
Plan the cutover like a surgery
Cutover is the weekend, or the days, when you stop using the old system and start using the new one. Everything before it is rehearsal.
What a cutover plan contains
- A freeze date after which no new master data or configuration changes go in without approval.
- A timeline by the hour for each task, with the person responsible and the person who verifies.
- Opening balances for the general ledger, accounts receivable and accounts payable, tied to a stated cut-off date.
- Open transactions, such as open sales orders, open purchase orders, and work in progress.
- An inventory count at or near the cut-off, since on-hand quantities are the figure most likely to be wrong.
- A go or no-go meeting with explicit criteria.
- A rollback plan that states what you will do if the new system cannot be used and by what point you decide.
Rehearse at least twice
Run mock conversions into a test environment, at least two times. Each rehearsal teaches you how long the data load takes, which records fail, and which reconciliations do not tie. Treat each mock as a graded exam. After the second mock, the failures should be few and understood.
Choose a cut-off that helps accounting
A start at the beginning of a fiscal period, ideally a month, keeps opening balances clean. Avoid cutting over during your busiest week, around quarter-end if possible, and in the middle of hurricane season if you have the choice, because a storm that takes down power and connectivity for several days is the worst possible time to learn a new system. If you operate in Puerto Rico, build a contingency for connectivity at the warehouse, such as a cellular backup for the scanners, and confirm what the cloud ERP vendor offers for offline operation.
Parallel run, and what it really means
A parallel run keeps the old system operating in step with the new one for a period, so you can compare results. It is expensive, because people do the work twice, so be deliberate about what you run in parallel.
A narrow, meaningful parallel run
For most small and midsize firms, I recommend a limited parallel.
- Run the first full month-end close in the new system, and reproduce key reports from the old one. Compare the trial balance, inventory valuation, receivables aging and payables aging.
- For a defined period, push a sample of orders through both systems, say one day per week for two weeks, and compare pricing, taxes, and totals.
- Count a sample of inventory locations daily, and compare the physical count to both systems.
Define tolerances before you start. A difference of a few cents from rounding may be acceptable. A difference in inventory value that nobody can explain is not.
When not to run a full parallel
Running two full systems for several months exhausts the team and often means neither is done well. If your mock conversions reconcile and your testing is solid, a short parallel on the highest-risk processes is enough. The alternative is a clean cutover with intensive support in the first two weeks.
The financial close is your final exam
The ERP project is truly done when you close the books in the new system and the numbers are accepted by finance, by your accountant, and by anyone who depends on your financial statements, such as a lender.
Plan for the close from the first week of the project.
- Involve the controller or bookkeeper in design, especially the chart of accounts, dimensions such as department or location, and the inventory costing method.
- Decide how you will handle costing. Standard cost, average cost and FIFO give different answers, and the choice affects margin reports. This is a decision for finance, with the accounting rules in your jurisdiction in mind. If in doubt, ask your accountant.
- Test the full period-end routine in the test environment. That means accruals, depreciation, inventory adjustments, intercompany eliminations if relevant, and period locking.
- Reconcile subledgers to the general ledger. Inventory, receivables and payables should each tie to a control account.
- Plan for a longer first close. A reasonable expectation is that the first close takes noticeably longer than the old one, and the second is faster.
A first-close checklist
- Bank accounts reconciled and matched to ledger.
- Receivables aging ties to the control account.
- Payables aging ties to the control account.
- Inventory valuation report ties to the inventory account.
- Goods received but not invoiced is reviewed and explained.
- Tax and payroll postings reviewed.
- Period locked and a copy of reports saved.
Training, testing and people
Technology is the easier part. The harder part is that a clerk who knew the old system by muscle memory must now relearn a day's work.
- Training should be role-based and use your own data, in the language people work in. In a bilingual team, plan training and quick-reference sheets in both English and Spanish.
- User acceptance testing is performed by the people who will use the system, against the scenarios from the fit-gap analysis. Write the expected result before the test, and sign off each scenario.
- Name super users in each department, people with enough time to help colleagues in the first weeks. Free up their schedule, since this is real work.
- Expect a productivity dip. Plan staffing and customer communication for it.
What to ask the vendor and the implementation partner
- Who will actually do the work, and what is their experience with companies like mine?
- What is in the fixed scope, and how are change requests priced?
- How many mock conversions are included?
- What is the support model after go-live, and how quickly do you respond to a stopped shipment?
- How do upgrades work, and what happens to my customizations?
- Can I export all of my data, in a documented format, at any time?
- Which parts of my project depend on your product's roadmap rather than what exists today?
Ask for references from firms of a similar size and process type, and speak with them yourself. Ask what they would do differently.
Risks worth naming early
| Risk | Early warning | Response |
|---|---|---|
| Dirty master data | Mock loads keep failing | Assign owners, delay go-live if needed |
| Scope growth | A "later" list that keeps shrinking | Require sponsor approval for every addition |
| Over-customization | Gaps outnumber fits | Revisit the process change option |
| Key person overload | Same three people in every meeting | Backfill their daily work |
| Weak cutover plan | No rollback criteria | Write them and rehearse |
| Connectivity loss | Scanners stop in an outage | Cellular backup and a manual fallback form |
Your first week
If the project has not started, take these steps this week. Write the purpose statement and three to five outcomes. Name the executive sponsor and the project lead. List every master data domain and assign an owner. Pull a sample of 100 item and 100 customer records and look at them honestly, since that sample tells you more about your timeline than any vendor estimate.
If you want help thinking through the plan, Ozcorp InfraOps works with small manufacturers and distributors on the data and integration side of projects like this, and InfraOps Base starts at $500 per month with vendor and license charges billed separately. I am glad to talk it over, and I will tell you plainly if a smaller step would serve you better.
Discuss your infrastructure requirements
Review the current service offer or schedule a conversation about the scope you need.
Stay in touch
Follow Ozcorp
Leave your details for occasional notes about new articles and Ozcorp updates.