Best Practices

What separates teams that get value from Intelligence from teams that stall.

Treat coverage as the foundation

Analysis cannot recover an observation that was never captured. When a result looks wrong, check coverage before questioning the analysis. Work the coverage cards on the Intelligence hub until each one is resolved rather than leaving them at needs confirmation indefinitely.

Keep profiles honest and current

Answer the interview as the operation actually runs today.
Republish when your role, sites or commercial terms change. A detention rate with a stale effective date quietly misleads.
Use custom instructions for terminology your team actually uses, so answers come back in your language.

Ask precisely

Name SiteTrax.io, the site, the period and the asset type. Most disappointing answers are underspecified questions rather than product limitations.

Verify anything that will be acted on

Before a charge is disputed, an invoice paid or a number published, open the underlying record and look at the image. The evidence trail is the reason to use SiteTrax.io rather than an estimate, so use it.

Use the language precisely

Say	Not
No observation of the unit in that period	The unit was not there
The reading could not be resolved	The container is damaged
The observed interval was 52 hours	The unit was delayed 52 hours
Stored readings	Units, when counting records

Give feedback on bad reads

Mark records correct or incorrect from the detail page. It takes a moment and it is the mechanism by which capture quality problems become visible.

Keep access tight

Review project users quarterly, including service accounts.
Grant the notifications write scope only where a workflow needs it.
Never put passwords or tokens in a prompt or a support email.
Connect assistants with the same account that published the profile.