Measurement and judgement
Automated checks can find repeatable facts quickly. The trouble starts when a long list or precise number is treated as a judgement about what matters most. GreatInternet uses automation to collect public evidence; a reviewer decides what the evidence means, how confident it is and what deserves attention first.
Useful measurement
Use it for facts that can be tested consistently: whether a link returns an error, metadata is present, how much a page transfers, how long a response takes, or whether a measurable attribute has a particular value. It is especially useful for repeat monitoring and regression checks.
Useful judgement
A repeated fact may matter greatly on one journey and barely at all on another. Judgement connects the observation to the visitor, the task they are trying to complete and the consequence of the friction.
Where the outputs differ
| Question | Automated score | Independent review |
|---|---|---|
| Starting point | A predefined set of checks. | A focused set of public pages and the journeys between them. |
| Main output | A score, pass/fail result or issue list. | An ordered view of what matters, what works and the next sensible move. |
| Uncertainty | Depends on what the tool measured and how it handled the page. | Confidence and limitations are visible; uncertain candidates can be withheld. |
| Patterns | Good at repeating the same check page by page. | Compares pages as a set and groups symptoms under a shared cause. |
| Strengths | Can show successful checks. | Names the strongest pattern and what should be left alone. |
| Cause | May suggest why a result occurred. | Separates the observed behaviour from any inferred cause. |
Directional, not decorative
GreatInternet uses Strong, Sound, Needs attention and At risk only when the evidence supports the view. Numerical precision belongs where a real measurement supports it—not where a score merely looks authoritative.
Read how findings are checked →The observation and its source stay attached.
The level explains how directly the evidence supports it.
Unknowns are named instead of quietly filled with certainty.
Working pages and patterns become benchmarks to preserve.
A useful review can prevent unnecessary change.
Priorities are ordered by consequence, not checklist volume.
Often the useful answer is both: measurement to collect the facts, judgement to decide what they mean.
Start with a free review of the public website. Go deeper only when the report is useful.