Evidence
A finding should
be checkable.
The useful question is not whether a claim sounds persuasive. It is what was checked, what the evidence supports and where it stops.
Keep the kinds of evidence distinct.
- Observed
- What a source, interface or record showed, with the relevant context and observation time.
- Reproduced
- Behaviour repeated through a recorded check, within its stated environment and scope.
- Inferred
- An interpretation, estimate or proposed next step. Its assumptions remain visible; it is not silently upgraded into a measured fact.
- Not established
- Missing access, conflicting definitions, stale information or an incomplete check. An absent result is not a clean result.
An interface is not the whole handoff.
In a UK wellness engagement, a sign-up form showed “Sent” while the server returned HTTP 403. The visible success message and the underlying response disagreed.
The homepage presents that finding as a typographic reconstruction, not an account screenshot. Other authorised examples are collected on Selected Work, with their evidence and current status.
Status is part of the evidence.
Recovered profiles are not revenue. Prepared campaigns are not sent campaigns. A deployed tool is not evidence of adoption or return. A development build and a research demonstration remain distinct from an operational service.
Private account records, client materials and community images are not republished merely because they would make a stronger-looking example. Public-source checks and illustrations are labelled for what they are.
Talk through one workflow.
Tell me what you are trying to improve, what you are still managing by hand, or what is not adding up. A few sentences is enough. No technical brief required.
An initial fit-check message does not require a paid consultation. Paid work is scoped and priced in writing before it starts.