Set it up so it can fail
The most common way a pilot goes wrong is that nobody defined what failure looks like, so it drifts into a permanent half-adoption. Before the first day, write down the end date, the workflow, and the two or three things that must be true for this to be a yes.
Something the business does every week, so the pilot gets three or four repetitions rather than one.
The people who do the work, not a project committee. Three to eight is the range where everyone is accountable.
Plus Chat, always — the pilot needs somewhere for questions.
This week's invoices, this week's leads, this quarter's project. Synthetic data hides the problems you are piloting to find.
A pilot without a date does not end, it fades.
erp.io pilot
Ends 31 March · 4 people
To do3
In progress2
Done4
What to measure
| Measure | How | A yes looks like |
|---|---|---|
| Did the work get done? | Compare the pilot week to the same week in the old system. | Same output, no heroics. |
| Did people use it unprompted? | Read the audit log and the activity feed. | Activity on days nobody was chasing. |
| How many workarounds appeared? | Count the times somebody went back to a spreadsheet. | One or two, each with a named reason. |
| What broke? | A written list, not a feeling. | Everything on it is either fixed, worked around, or accepted. |
| Would the team choose it? | Ask them individually, not in the review meeting. | They would. |
When the pilot succeeds, do not start again in a clean organisation. The pilot organisation becomes the real one — the settings are already right and the pilot team are already the people who know how it is configured.
The two failure modes
Too many modules. A pilot that enables everything measures nothing, because attention splits and no single workflow gets four repetitions. Enable what the workflow needs.
Too little authority. If the pilot team cannot change the chart of accounts, the pipeline stages, or the project structure without asking, they will conform their work to the defaults and you will learn nothing about fit. Give them an administrator between them.
What this does not do
A pilot runs on the ordinary trial. If it needs to run longer than 30 days, ask.
There is no separate test tenant. A disposable second organisation is the closest equivalent.
Turning a module off retains its data but there is no snapshot-and-restore.
Questions
Can a pilot run alongside our current system?
For the pilot workflow, yes, and it usually should. Full parallel running across the company is a different and heavier thing.
What if the pilot is a no?
Then it cost you three weeks instead of nine months, which was the point. Write down why — it is the most useful artefact the pilot produces.
Related
The order that makes a migration survivable: structure first, opening balances second, transactional history last — and the checks that prove it worked.
Your first weekA day-by-day plan for the first week in a new erp.io workspace: what to set up, what to leave alone, and what to check before you trust it.