Alerts

• Coverage gapA requirement that applies to you has no proof behind it, or only part of what it asks for. One alert per requirement, naming the requirement and the duty it belongs to. What to do: record evidence.
• Evidence out of dateA requirement you had satisfied no longer is, because the proof has aged out of the window it had to be repeated in, or is tied to a superseded version of your record. Deliberately a separate kind from a coverage gap, because the work differs — you are refreshing something rather than producing it for the first time. Usually the highest-value item in the inbox.
• Rule pack updatedThe rules Clause50 evaluates against have moved on since your last calculation. The alert carries the difference between the two versions, narrowed to the rules that actually affect this system, so you are not reading changes irrelevant to you. The hourly pass moves you onto the new rules on its own; the alert exists so you know your obligations changed, not so you can press a button. See when the score updates.
• System changedA new version of the system record was opened after coverage was last calculated, so the figure on screen described an older description of the system. The next hourly pass rescores it against the new version; the alert is there so a version bump is visible to everyone watching, not only to whoever made it.

A system whose coverage has never been calculated produces no alerts at all — not even gap alerts. Every one of the four kinds is a comparison against a previous calculation, and there is no honest baseline to compare against. The dashboard already tells you the system has not been checked yet; piling alerts derived from nothing on top would imply a baseline that does not exist. In practice this state lasts less than an hour: the same pass that evaluates alerts calculates coverage first, so a system with no baseline gets one and starts producing alerts on the following pass.

Each alert is a dated statement: at this time, this system was in this condition. It is a record of what you were told and when. Once written it is not edited — the only thing that ever changes on it is whether the email went out.

A condition that clears simply stops producing new alerts. Record the missing evidence and no further gap alert is raised for that requirement. The old one stays in the inbox as history, which is the useful behaviour: the question an auditor asks is not “is this ticket closed” but “when did you know, and what did you do about it”.

So the inbox is a timeline, and the coverage page is the current picture. Use each for what it is good at.

Clause50 runs a scheduled pass every hour. Each pass does three things in order: recalculates the coverage of any system whose score has fallen behind, re-examines alerts for every system against that fresh result, then sends each organisation with anything new one digest email listing everything not yet notified. Never one email per alert.

Hourly is the cadence of the check, not of your inbox. What stops you being emailed every hour is the suppression rule below, not the schedule — a condition you have not dealt with is raised once and then left alone for a week.

• One email per organisationIt goes to the email addresses of the people in your organisation.
• The same condition does not nagAn alert is suppressed if the same condition is already waiting to be sent, or was raised within the last seven days. A gap you have not got to yet does not generate a fresh email on every pass, or every morning.
• The email says less than the inboxThe digest is assembled from a fixed, narrow set of fields — the system name, the requirement, the rule reference, the versions. It never copies the full contents of an alert into an email.
• Nothing is marked sent unless it was sentIf email delivery is unavailable, the alerts stay in the inbox marked Not sent yet and are retried on the next pass. Clause50 does not mark something notified that nobody was notified about.

You do not have to wait for the next pass. There is an evaluate now action per system that updates the inbox immediately; it does not send email, which follows on the next scheduled pass.

Nothing is marked sent unless it was sent — see below — so an alert raised while email delivery is unavailable is not lost. It waits in the inbox and goes out on a later pass.

Alert records and digest emails carry identifiers, names, statuses, and rule references. They never carry the contents of your evidence, your uploaded files, or any connector credential. That restriction is enforced where the email is assembled, not merely intended where the alert is written.

Automatic alerts are listed as a paid-plan feature in the plan comparison on the Artifacts page.