Implementation craft

Everything up to here was about Odoo. This stage is about projects, and it is the part that decides whether you are hired again.

Consultants who know the product but not this are the ones whose projects go late.

Discovery

The skill is interviewing, not note-taking.

• Talk to the people who do the work, not only their managers. The manager describes the official process; the clerk describes the real one, including the spreadsheet that holds it together.
• Ask for numbers. Orders per day, SKUs, users, invoices per month. Volume changes design.
• Ask what goes wrong. “What is the most annoying part of your week?” finds more requirements than any structured questionnaire.
• Watch, if you can. Sit beside someone for an hour. You will see steps nobody would ever have mentioned.
• Find the informal system. The WhatsApp group, the notebook, the shared file. It is always load-bearing.

Write it back to them and have them correct it. What people confirm in writing is much more reliable than what they said in a meeting.

Fit-gap

Every requirement gets one of four verdicts — standard, configuration, custom, or process change — with days attached. This is the document that becomes the scope, and producing it well is the most valuable thing you do.

Two habits:

Push hard toward standard. Every custom line is a permanent cost. If a process change avoids it, recommend the process change and say plainly what it will cost the client in habit and training.

Estimate honestly, including testing. A beginner estimates the coding. Add analysis, testing, fixing, documentation and the meeting where it is demonstrated. Being trusted is worth more than being cheap, and an estimate you meet is worth more than one that wins the argument.

Data migration

The phase that most often runs late, because everyone underestimates how bad the client’s data is.

• Ask for a real extract in week one. Not a sample — the whole thing. It is always worse than described.
• Migrate master data first, then open items, then history. Most history is not worth carrying; ask what it will actually be used for.
• Script every import so it can be re-run. You will run it many times. Manual edits during import cannot be repeated and cannot be trusted.
• Reconcile with the client’s accountant, not with yourself. Opening balances they have not agreed are opening balances that will be disputed.
• Cleaning is the client’s job. You can make it easy; you cannot decide which of two customers is the real one.

Testing and training

Write test scripts from the acceptance criteria, per role, in the order someone really works. Have the actual users run them on migrated data.

For training: train roles, not the software. A storekeeper does not need to see the accounting screens. Record the sessions — clients ask for them constantly, and it saves you from repeating the same session for every new joiner.

Expect some “this is wrong” to mean “this is different”. Learn to tell the two apart without being dismissive.

Go-live

Boring is the goal. A cut-over plan with a named owner and a time for each step, a freeze on new work weeks before, a rollback that has been tried, and a date chosen for the client’s calendar rather than yours.

Then be present. The first three days shape how the client feels about the entire project, and about you.

Hypercare

Support through the first full close, when the real problems appear — a tax posting somewhere unexpected, a valuation that does not match, a report the auditor wants in a different shape.

Then hand over deliberately: documentation, training material, a runbook, and a written list of the compromises made. A clean handover is what gets you recommended.

The habits that matter

• Write things down. Decisions, agreements, compromises. Memory is not a project record.
• Say bad news early. A slip reported in week two is a schedule adjustment. The same slip reported in week nine is a crisis and a lost client.
• Never say yes in the room to something you have not estimated. “Let me check and come back tomorrow” is always acceptable.
• Keep a decision log. It ends most disputes in a minute.
• Demonstrate early and often. Clients cannot review a description; they can review a screen.