# Test-Set Sheet

Use it before you run anything: it fixes the jobs you will test your workflow on, and what a good result looks like for each.

## About the workflow

- Workflow name:
- What it is for, in one sentence:
- Who reads the result:
- Version of the workflow (a date or a number):
- Tools you will test it in:

## The checks you will score

One row for each check. Say who runs it: a rule you can count, an AI second look, or you. Write both sides, so two people would mark it the same way.

| Check | Who runs it | Met when | Not met when |
| ----- | ----------- | -------- | ------------ |
|       |             |          |              |
|       |             |          |              |
|       |             |          |              |
|       |             |          |              |
|       |             |          |              |

## The test set

Three to five jobs. Include at least one typical job, one easy job and one tricky job (a real complication, such as a name that looks like another name). Keep the source material beside each, unchanged.

| Case | Kind | The job | Where the source is kept | What good looks like |
| ---- | ---- | ------- | ------------------------ | -------------------- |
|      |      |         |                          |                      |
|      |      |         |                          |                      |
|      |      |         |                          |                      |
|      |      |         |                          |                      |
|      |      |         |                          |                      |

## Rules for a fair test

1. Change one thing at a time. Same input, same checks, same version.
2. Keep the cases that fail. Run them every time.
3. Write what good looks like before you run, not after.
4. Count the same rows every time. If a criterion was wrong, change it, say so, and re-score the earlier runs.
5. Leave judgment checks to a person, and record them as your read.
6. After a fix, run the whole test set again.

## Next

Run the first case, then fill in a Run-Record Sheet for it.

---

From Logic Lab (logiclabhq.com). Free to use and adapt, with credit.