The four levels
| Level | What it means |
|---|---|
| Critical | The problem stops something that can't be worked around, in one of five ways (below). |
| High | People can still finish, but many won't: they take a wrong turn, lose confidence or give up. |
| Medium | Friction. People notice, work around it and carry on. |
| Low | Worth fixing, with no measurable cost to the task. |
What makes an issue critical
An issue is critical only when its worst case does one of these five things. If none applies, it's high at most.
- Blocks the task. Nobody can finish buying, signing up, signing in or sending the form, and there's no way around it on the page.
- Shuts a group out. Keyboard-only or screen-reader users, or a screen size, can't use it at all.
- Carries legal exposure. Consent and tracking, a required price or delivery disclosure, misleading commercial practice.
- Costs money or data for good. An unexpected charge, lost form data, a destructive action with no confirmation and no undo.
- Puts someone's safety at risk. For example, a support service for people facing abuse that offers no quick way to leave the page.
Why there's no score
A score adds ten small problems up until they look as bad as the one that breaks checkout. Severity keeps them apart. You see the blocker on its own at the top, and the polish further down.
What each issue includes
Four parts, the same as in every Rubric report:
- what's wrong;
- why it matters;
- what to change;
- the research behind it.
The report ends in your fix list: one task per problem, worst first. Download it as HTML or PDF.
Examples
| Issue | Level |
|---|---|
| Checkout needs an account, with no guest option | Critical: blocks the task |
| The total with shipping and tax never appears before payment | Critical: legal exposure |
| The main navigation can't be reached by keyboard | Critical: shuts a group out |
| Weak or missing trust signals at checkout | High: costs conversion, blocks nothing |
| More than 20 top-level categories | Medium: friction |
Good to know
Who decides the severity?
Each check has a level, set once against the five tests above and reviewed for the whole library. A finding takes the level of the check it fails, so the same problem gets the same level on every site.
Can I see the checks?
Yes. The guides at rubric.design/ux show each rule, how it usually fails and how serious it tends to be.
How many pages?
Up to seven per audit. Rubric picks the pages that matter, or you choose.
What language are reports in?
English.
