The decision that shapes everything: where you start
Do not start with the system that hurts most. Start with the one that has the fewest dependencies.
| If you start with… | Advantage | Risk |
|---|---|---|
| Maintenance | A contained area, few people and visible results in a month | Low |
| Support | It comes in through channels and connects easily | Medium: it is your face to the customer |
| Sales | Few people | Low, but it adds little at first |
| Connecting the ERP | None as a starting point | High: never first |
An order that tends to work: maintenance, field service, support, sales and, last, the connection with the ERP that does your invoicing. Each phase feeds the next, and billable work reaches the ERP when the data it depends on is already reliable.
The coexistence rule
During the transition you will have two live systems. The rule that prevents disaster: each piece of data has one system in charge of it.
If customers are created in the new system, they are created only there, and synced to the old one. Never in both. The day someone creates a customer in both places, the records start to drift apart, and that never gets fixed.
Phase by phase
Phase 0. Inventory of the truth (one week)
A list of which data lives in which system and which system is in charge of it. It is boring and essential, and it is where the surprises appear: the three systems usually have different customer lists.
Phase 1. First module (three or four weeks)
Set it up with real data, test it with a small team and switch over. A module left half-done indefinitely is the worst thing that can happen.
Phase 2. Second module (three or four weeks)
This is where the first real benefit shows up: two areas sharing the same customer record.
Phase 3. Channels
Email, WhatsApp and phone move to the new system. It is the point of no return, and it is best done when the team is already comfortable.
Phase 4. Connecting the ERP
Last. Invoicing and accounting stay in your ERP: what changes is that billable work reaches it without retyping. By then, hours and parts are coming in properly, and that is what makes the invoice reliable.
How to import without duplicates
- Export what is there first, as a backup.
- Do a dry run. Validate the whole file without saving anything, and read what it would create, what it would update and what would fail.
- Test the real import with ten rows and check the result by hand.
- Import the rest and fix only the rows that fail, not the whole file.
- Pick a code that identifies each record, such as each customer’s tax ID, and never change it. With it, you can repeat the import without creating duplicates.
And if you only need to fix one field, such as the account owner of each customer, you do not need to reimport everything: upload the code and that one column.
What almost always goes wrong
1. Duplicates show up in phase 0 and nobody wants to decide. Three customer lists with different names for the same company. Someone has to decide which one wins and clean it up, and they need the authority to do it.
2. Someone keeps using the old system. There is always someone who finds it “handier there.” If you do not truly switch off, they contaminate the data for months.
3. The old system gets copied as it is. Flaws and never-used fields included. Migration is your only chance to clean up, and wasting it is expensive.
4. People outside the project are not trained. The pilot team knows everything, and everyone else finds out on the Monday it goes live.
What not to migrate
- History older than 24 months. It slows things down, adds clutter and nobody looks at it. If needed, keep an export.
- Fields nobody fills in. If 90% of a field is empty in the old system, leave it behind.
- Settings for a process that no longer exists. Ask “why is this here?” If the answer is “I don’t know,” do not migrate it.
How to know it worked
Three months after the last switch-over: nobody opens the old system, no data is typed twice and the reports that used to be built by hand simply open.
It is not a weekend project, and it is not a year-long one either. It is four or five phases of a few weeks, each ending with a real switch-over. And phase 0, the boring one, decides whether it works.
Last updated: