Test a Workflow Before Switching It On: A Checklist
How to test a workflow before switching it on: run it end to end on a contact you control, cover every branch, read it on a phone, then check the records.
To test a workflow before switching it on, run it end to end on a contact you control, ideally your own, check every branch, read the messages on a phone and then inspect the records it left behind. A test takes minutes and catches the errors that would otherwise reach your best prospects first.
Most automated embarrassment is preventable, and the cheapest prevention is looking at the thing before anyone else does. This post is a pre-launch routine. It assumes you have already chosen the workflow and designed it, and it covers the last gate before real people are involved.
How to test a workflow before switching it on: the principle
Run the whole workflow, on real data, on a person you control. Not a part of it, not a preview, not a reading of the plan. The whole thing, from the trigger to the last step, so that what you see is what a recipient would see.
There is one fact to hold in mind before you begin. In many tools a test run is a real run, and real messages go to the contact you choose. That is exactly why the contact should be you. Check how your own tool behaves, and never assume that a test is a rehearsal.
| What you test | Why it needs doing |
|---|---|
| The trigger fires | A rule that never starts looks the same as a quiet week |
| Every branch | The happy path is the one you already believe in |
| The messages | Links, names and layout fail in ways a plan cannot show |
| The timing | Waits and arrival hours change how a message feels |
| The records afterwards | A step can run and still do nothing useful |
Step one: use yourself as the contact
Create or choose a contact that is you, and enter it into the workflow the way a real person would arrive. If the trigger is a form, submit the form. If it is a stage change, move a record. The aim is to fire the real trigger, not to push yourself into the middle.
Then watch. What arrives, and when? What does it look like? The first run often shows something you did not design: a message an hour later than you pictured, a greeting with an empty gap, a task assigned to nobody.
Step two: walk every branch
A workflow with forks has more than one story, and each story needs its own run. Do not stop when the main path works. Set up your test contact so that it takes the other path, then run it again. If a condition depends on a field, test with the field filled in, with it empty and, where it matters, with it filled in badly.
The reason is the one explained in how to set workflow entry criteria: a condition with only a yes path is a place where people silently fall out. A run down the no path shows whether the person ends up somewhere sensible.
Step three: read it as the recipient would
Open the message on a phone, because that is where many people will meet it. Look at the greeting, the links, the layout and the sign-off. Click every link. Then read it for tone, asking whether it pretends to be a personal note from someone who did not write it.
The checklist for content is short.
- Every personalised field reads naturally when it is empty or badly formatted.
- Every link in every message works.
- No message pretends to be a personal note from someone who did not write it.
- Every marketing message carries a working way to stop receiving it.
The last item deserves emphasis. Anyone who receives messages from your automation must be able to stop receiving them, easily, and the automation must honour that without exception. That includes sends from workflows that were built before the person opted out. Getting it wrong is embarrassing, and in many places it is unlawful.
Step four: check the records afterwards
The messages are the visible half. The invisible half is what the workflow did to your records, and it is where silent errors hide. Open the contact you used and look.
| Look for | Question to ask |
|---|---|
| Tasks | Was it created, and does it name the right person? |
| Tags | Was it applied, and only once? |
| Owners | Is there one owner, and is it the intended one? |
| Notifications | Did the right teammate hear about it? |
| The timeline | Does the history show what happened, in order? |
If a step ran but its effect is missing, you have found a silent failure before it could matter. The defence against the same failure after launch is in how to spot a silently failing workflow.
Step five: decide duplicates and timing on purpose
Some questions cannot be answered by one run. They need a decision. Settle them before switching on.
- Can a person enter this workflow more than once? By default a person already inside should not be enrolled again. Allow re-entry only where it is clearly intended, such as a workflow that should run every time a customer renews.
- When do messages land? A message in the middle of the night or on a public holiday announces that nobody was there to send it. Make waits respect working hours where you can.
- What happens if the situation changes during a wait? The goal that ends the workflow should remove the person whatever step they are on.
- Does any other workflow send this person something conflicting at the same time?
Duplicate messages tell the recipient, very precisely, that no human is paying attention. A test can show a duplicate if you deliberately try to enter the same contact twice, which is worth doing once.
Test the exit as well as the entry
Most people test that the workflow starts. Fewer test that it stops. Try it: bring your test contact to the goal that should end the workflow, such as replying, booking or buying, and confirm that the person leaves, whatever step they were on. Without this, the automation keeps talking after the conversation has moved on, and the reminder to book a call goes to someone who booked yesterday.
This is the exit rule written into the design, now proved by a run. The thinking behind it appears in how to decide what to automate first.
A pre-launch checklist
- I ran the whole workflow on a contact I control, knowing the test sends for real.
- I tested every branch, not only the happy path.
- I read the messages on a phone, as the recipient would.
- I checked the records afterwards: tasks, tags, owners and notifications are where they should be.
- I decided whether a person can enter more than once, on purpose.
- Messages land during sensible hours.
- Reaching the goal removes the person.
- No other workflow sends a conflicting message at the same moment.
What a test usually reveals
The errors a test catches are rarely exotic. They are the ordinary ones that a plan hides, and knowing them in advance helps you look in the right places.
| What you see | What it usually means | What to do |
|---|---|---|
| A greeting with a gap in it | A personalised field was empty | Rephrase so it reads naturally without it |
| A link that goes nowhere | The address was mistyped or moved | Fix it and click every link again |
| Everyone takes the same path | A condition reads a field that is empty or unreliable | Check the field in real records |
| Nothing arrives at all | The trigger did not fire for the way you entered | Check what really starts the workflow |
| A message arrives at an odd hour | The wait ignores working hours | Change the timing so it respects them |
| The task or owner is missing | A step ran without its effect | Treat it as a failure, not a success |
Each row corresponds to something that would otherwise have reached a real prospect first. That is the whole value of the exercise, and it is why the routine is worth repeating every time rather than once.
What a test cannot tell you
A test proves that the workflow does what you designed. It cannot prove that you designed the right thing. It also cannot show you the people who arrive by routes you did not anticipate, or the fields that are empty for other contacts. Those only appear once real people are inside.
So the test is the start of watching, not the end of it. Read what the workflow does in its first days, and keep reading. A test is a gate, and the review after launch is the loop that keeps it honest.
Frequently asked questions
Why use my own contact and not a colleague's?
Because the test may send real messages. Using yourself means any mistake lands on a person who understands what happened.
Is testing the happy path enough for a simple workflow?
No. Even a short workflow has an alternative path whenever it has a condition, and the quiet path is where people fall out unseen.
How long should a test take?
Minutes, for most workflows. That small cost is the reason to test every time, not only when you feel unsure.
Do I need to retest after a change?
Yes. Test again after any change to the workflow, and after any change to forms, fields, stages or pipelines that it depends on.
The full method, with the checklists, is in The Automation Playbook, a free 29-page playbook.