Support
The SLA clock
Business minutes, not wall-clock — counted against the same working patterns and holiday calendars the rest of the product already uses.
Every desk has tickets. What separates them is whether the clock is believable. This one is built to be argued with: nothing is stored, everything is recomputed from what actually happened, and every figure can be traced to an event.
It counts working time, not elapsed time
A ticket raised at five on Friday evening has not missed a four-hour target by Saturday lunchtime. Opening hours come from a working pattern — the same model that says when a person works — and the public holidays compose through the same rule leave uses: federal plus state plus company, unioned and deduplicated.
That last part is the piece a standalone help desk cannot do. Its business calendar is a list somebody typed in and will not maintain; here it is the same calendar your leave and your sprint capacity already run on, so it is right for the same reason those are.
Nothing is stored
There is no countdown in a column anywhere. Each time a ticket is read, its history is replayed and the number worked out again. A stored countdown would be wrong the moment a policy, a calendar or a state changed — and wrong invisibly, which is worse than wrong.
Which is why the event log exists from the start
Every state change is recorded as it happens. A clock added later to tickets whose history was never kept could only start counting from the day it was switched on, and every older ticket would report a number nobody could defend.
What a badge means
| Badge | What it is telling you |
|---|---|
| Nothing at all | No target was promised for this ticket. Not the same as a target that was met. |
| Met in 40m | Answered inside the promise, and how long it took in working time. |
| 2h left | Still running. |
| 30m left, in amber | At risk — four fifths of the budget is gone. A proportion, not a fixed margin, so long targets stay useful. |
| Paused, 3h left | Waiting on the customer. Not a warning, because nobody here is failing to act. |
| Missed by 20m | Answered, late. |
| Overdue by 3h | Still not answered. |
Missed and overdue are kept apart on purpose. One is a post-mortem; the other is a customer still sitting there. A desk that shows one badge for both cannot tell anybody which it is.
Which promise applies
SLA policies are scoped by customer, priority and channel, and the most specific one wins — the same rule leave policy, overtime rules and contact rules all follow. A policy the customer bought beats a general one about their priority.
A policy scoped elsewhere is never a fallback
A policy for urgent tickets does not quietly cover a normal one because nothing else matched. Where nothing applies, the desk says no target was promised — which is a real answer. Inventing one would manufacture a breach out of a promise nobody gave.
The policy is resolved when the ticket is raised and kept with it, so editing a policy next month does not silently restate what was promised on a ticket already answered. The clock stays derived; only the promise it is measured against is fixed.
Leave, and what it is actually good for here
The clock does not pause because the agent is on holiday
It is tempting, and it is wrong: an SLA is a promise to the customer, and the customer does not care who is away. A desk that stopped the clock every time somebody took a Friday off would report numbers that mean nothing.
What knowing about leave is good for is different, and it is something no separate help desk can do: not handing a four-hour target to somebody who is out until Thursday, and flagging tickets already assigned to somebody away whose target expires before they are back. Nobody is doing anything wrong in that second case — the ticket is simply going to be missed, and somebody can move it now.
Both are live: new tickets skip whoever is out when they are routed, and a warning lists tickets whose target expires before their agent returns. The second is narrow on purpose — a ticket they will be back in time for is not stranded, and a register where everything is flagged is one nobody reads.