Method
How the exit test is scored
The five checks, what each one is worth, and the thresholds that separate the readings in the free exit test.
- Rubric version
- v1.0.0
- Published
- 2026-09-03
- Last reviewed
- 2026-09-13
- Next review due
- 2026-12-13
What is actually being scored
This scores your evidence about a product, not the product itself. A strong reading means you tested the export and wrote down what happened; a weak one usually means nobody has run it yet. Software Sausage names and scores no vendor here — the product name and every answer stay in your browser and are never sent to a server. A share link carries them after the # in the URL, which browsers do not transmit, so they reach only the people you send the link to.
The five checks and their weights
Each check scores 0 to 3 points, multiplied by its weight. The maximum weighted score is 30.
| Step | Check | Weight | The question |
|---|---|---|---|
| 1 | Export | ×1 | Can you produce the export yourself, on the plan you would actually buy? |
| 2 | Open | ×2 | Opened without the vendor, does the artifact carry the working state or only records? |
| 3 | Restore | ×3 | Is the format documented, and is there somewhere specific it can be re-imported? |
| 4 | Operate | ×1 | After the last invoice, what still runs and who can still reach the data? |
| 5 | Reconcile | ×3 | Did you import it into a named alternative and reconcile the totals? |
How a reading is decided
- Untested — any check left unanswered. There is no verdict to give, and this takes priority over the score.
- Verified — no check scores below 2, and restore and reconcile must both score full points. A migration you have not restored and reconciled is not verified, however good the export looked.
- No usable export — a weighted score below 10.
- Archive — everything else: the data can leave, but the working state is not proven to leave with it.
Working migration path, verified.
You have moved this data into a named alternative and reconciled it. That is evidence rather than a promise, and it is rare. Re-run the test before each renewal: export behaviour changes with plans.
Archive, not a migration path.
Data can leave. The working state — relationships, history, permissions, and the settings that make the records mean something — is not proven to leave with it. Budget the reconstruction as labour, not as a free export button.
No usable export.
On the evidence you have, the working data cannot leave in a form anything else can read. That is a term to negotiate before signing, not a problem to discover at renewal.
Untested.
At least one check has no answer, so there is no verdict to give yet. An untested export is the most common failure in this whole test: the button exists, nobody has ever pressed it, and the first press happens during the migration.
Where the weights come from
They are judgement, not measurement, and the reasoning is the whole defence: a restore you have performed and a reconciliation you have checked are evidence, while an export button you have never pressed is a promise. The rubric weights evidence above availability for that reason. No dataset sits behind these numbers, and none is claimed.
Changelog
| 2026-09-13 | Reviewed. No change to any weight, threshold, or option value. Added this page and a version stamp so a score can be cited against the rubric that produced it. |
|---|---|
| 2026-09-03 | Rubric v1.0.0 published: five weighted checks scored 0 to 3, with the verified, archive, and no-usable-export readings. |
Read this beside the site methodology and the lifecycle cost method.