Test this for real
On some findings you will see a button that says Would this change help?
That is this feature. The name below is what we call it in the docs and the CLI.
We apply a temporary fix in our lab and measure again. You do not need access to the code. Costs 1 credit.
When the button shows
Today it shows for:
- Lazy load on the main image
- A blocking script or stylesheet we can rewrite
- A slow-filling page held up by a script we can name
- Late discovery of the main image (closest lab step)
- Scripts keeping the page too busy to answer taps, when we can name one
For the last one we delay the script the report named and measure again. That is the change the report told you to make, so it is the one we test. Delaying moves the work rather than removing it, so a flat result is a real answer: it means that change will not move this number, and you have saved the afternoon.
If the button is missing, we cannot lab test that finding yet, and you are told before anything is charged rather than after.
Two findings can look the same and only one gets the button. What decides it is
whether the report names the file. "Add defer to app.js" is something we can
apply and measure. "No single script explains this wait" is us saying we could
not pin the work on anything, so there is nothing to apply. Charging for the
second would be selling a measurement we cannot make.
The three answers
| Verdict | Meaning |
|---|---|
| It worked | The page got clearly faster than it moves on its own |
| It did not help | The fix did not make things better |
| We cannot tell yet | The change was smaller than the amount the page moves on its own |
If it fails
You get the credit back. A dead browser, a timeout, or a bot check on the second visit are our problem, not yours.
We cannot tell yet is not a failure and is not refunded. The measurement ran and that is the answer it gave.
Also on CLI and MCP
- CLI:
pnpm pafcore test 1 - MCP:
test_fix