Understand your evidence

CiteLadder shows saved observations. An audit captures a set of answers; a crawl captures website pages; an integration imports a particular provider report. Each describes what was observed under those conditions.

Zero is a result; unavailable is a limitation

State	How to read it	What to do
Observed zero	The eligible evidence was measured and the count was zero	Inspect the measured scope
Not run	This measurement has not been performed	Start an appropriate run if needed
Partial	Some evidence or checks are missing	Read the coverage and omissions
Unknown or unavailable	The evidence cannot support the requested value	Resolve the stated limitation
Failed	The attempt did not complete successfully	Inspect the failure and recovery action
Not applicable	The check does not apply to this case	Do not treat it as a failure
Historical	The evidence belongs to an earlier observation	Keep its original date and context

A failed AI answer is not an answer that omitted your brand. A missing traffic report is not zero traffic. These distinctions matter most when reporting percentages.

Follow the denominator

Ask what each rate is divided by. A brand appearance rate, a citation rate and a product’s share of shelf describe different populations.

Read the scope shown with the metric. Changing a table page does not change the full dataset’s totals. Do not reconstruct a headline value by adding visible rows together.

Compare like with like

Before describing movement, check the prompt set, engine, market, time window and relevant measurement conditions. A broader prompt portfolio can change a rate even if the answers to the original questions did not change.

If CiteLadder marks a comparison as unavailable or non-comparable, report that limitation. Choosing an unrelated period and calling it the requested comparison hides the problem.

Separate observations from explanations

“An answer cited this URL” is an observation. “That URL caused the brand to appear” is a causal claim that the observation does not establish.

The same applies after a website edit. Later evidence may show the expected change, or a metric may move, without proving that the edit caused every downstream result.

Use a repeatable reporting format

For each finding, capture:

• What happened: the specific observed result.
• Where and when: project, target, source and date window.
• What supports it: the answer, page or saved record.
• What is missing: coverage, incompatible scope or unresolved evidence.
• What comes next: a review, an Action or a later measurement.

This gives your team something it can verify and act on.