Method
Every line says how we know it
The skill labels each finding. It is the single habit that stops invented numbers reaching a client or a web page.

Which source wins when two disagree
Top beats bottom. State the conflict and why one source won.
What the skill will never invent
Rankings, traffic, search volume, keyword difficulty, backlinks, conversion rates, revenue, client names, quotes, results, certifications, awards and event dates. A missing figure is written "Not available" with the method that would find it.
When sources disagree
- The business's own records
- Government sources
- The event organiser or venue
- Search-engine documentation
- Primary research
- Industry press, then news
- Directories, then competitor sites
- Search snippets (never alone)
BreakdownStep by step
- Label every factual line
[V] verified from a live page or first-party record; [T] a tool observation; [A] analysis; [AS] an assumption stated openly; [NV] not verified and never published as fact.
- Attach the source
Each evidence row carries the exact URL or tool screen. "Search Console said so" is not enough; "Search Console, Pages report, 18 Nov" is.
- Rank sources when they disagree
First-party records beat government sources, which beat organisers, search engine documentation, research, press, directories and competitor sites. Say which won and why.
- Write "Not available" when you do not know
A missing number is honest. An invented one ends up in a sales deck and then in front of a client.
Worked example
Two ways to report the same finding
Weak: "The site is slow." Strong: "[T] PageSpeed Insights, mobile, three runs on 18 Nov: LCP 3.1, 3.4 and 3.2 s. Field data: not enough Chrome users yet. [A] The hero video loads before the headline; deferring it is the first fix to try."
TipsWhat works in practice
AvoidCommon mistakes
- Quoting search volumes from memory.
- Copying a competitor's claim as if it were a market fact.
- Mixing an old audit number into a new report without re-checking it.
Quest · Auditor badge
Label the truth
A number came from a single Lighthouse run. How should it be reported?
Why: Lab scores vary between runs. Report the spread of at least three comparable runs and keep field data separate.