Testing
How to Test One Flow in Your App Before You Fix Anything
There is a point in most app projects where you know something is off but not what. The sign-up feels slow. A friend said checkout did something strange. You could start fixing things at random, and with an AI tool it is tempting, because each fix takes a minute. The trouble is that random fixes move problems around.
A better first move is to test one flow, end to end, and write down exactly what happens. One journey is small enough to finish in an hour and clear enough to act on. This guide shows you how.
What a flow is
A flow is one path a person takes to get one result. "Sign up and reach the dashboard" is a flow. "Add an item, pay, and receive the confirmation" is a flow. "Forgot my password and got back in" is a flow.
A flow has a start, a finish, and a handful of steps in between. If you cannot name the start and the finish, you do not have a flow yet, you have a feature. Rewrite it until you can.
Step 1: Choose the flow that costs you the most when it breaks
Ask one question: if only one thing in my app works, which should it be? For a store, it is paying. For a membership app, it is signing up and signing back in. For a booking tool, it is making a booking.
That is the flow to test first. If it works, you have a base to build on. If it fails, you have found the most useful problem in your whole app.
Step 2: Write the steps before you run them
Write the journey as numbered steps, the way a stranger would describe it:
1. Open the link from an email on a phone.
2. Tap Sign Up.
3. Enter an email and a password.
4. Confirm the email.
5. Land on the first screen.
Writing the steps first stops you from skipping the boring ones. Step 4 is the step most people forget, and it is where a lot of sign-ups quietly fail.
Step 3: Run it on desktop, then on a phone
Run the whole flow on your computer first. Then run it again on a real phone, from the very beginning, not from where you left off. The phone run is where layout problems show up: buttons that are cut off, a keyboard that hides the field, a page that needs sideways scrolling.
At each step, note two things: what you expected and what happened. When they match, write nothing. When they do not, write it down at once. You will not remember the details later.
Step 4: Record each problem so it can be repeated
For every problem, write:
Where you were (the step number).
What you did.
What you expected.
What happened instead.
The device and browser.
"Checkout is broken" gives nobody anything to fix. "Step 6, on a phone, tapped Pay with a saved card and the screen went blank; I expected a confirmation page" gives you, or your AI tool, a clear target. If you can, add a screenshot.
Step 5: Rank the problems by what they cost
Not every problem matters equally. Rank each one:
Blocks the result: the person cannot finish the flow.
Slows the result: they can finish, but it is confusing or hard.
Polish: it looks off but does not get in the way.
Fix the blockers first, in the order people would hit them. A blocker at step two hides everything after it, so the problems at steps three to six may disappear once step two works.
Step 6: Fix one thing, then run the flow again
This is the habit that saves the most time. Fix one problem, then run the whole flow again from the top. AI tools are fast, and a fast tool can break a working step while it fixes a broken one. Running the flow again is how you catch that.
When the flow runs clean on both desktop and phone, you can move to the next one.
What to do when you find nothing
If the flow works cleanly, try again with a worse day. Use a slow connection, a second account, a different browser, and one wrong entry on purpose. A flow that survives that is in good shape.
Related reading: How to Test Your AI-Built App Before Launch for the full plan once your first flow works.
If you want one journey tested end to end on desktop and mobile, with steps to repeat each issue and a fix direction for each, the One Flow Test does exactly that. See the One Flow Test