Roster, coverage and redistributing work
Your queues already have opening hours, and your people already say whether they are available. What the product never knew was the third thing: **who is supposed to be covering which queue, and when**. Without that plan there is no adherence, no occupancy, and no answer to "the queue opens at 8am t
What this page answers
Your queues already have opening hours, and your people already say whether they are available. What the product never knew was the third thing: **who is supposed to be covering which queue, and when**. Without that plan there is no adherence, no occupancy, and no answer to "the queue opens at 8am tomorrow and is anyone on?".
The roster is that plan. Open it in Settings, under Service, in **Roster**.
What a roster is made of
- **The roster** points at one queue (or at nobody, when it is a general on-call plan) and carries its own **time zone**. A team in another country reads its own clock, and the grid only makes sense that way. - **Shifts** repeat weekly: a weekday, a start, an end, and how many people the operation wants on. An end past midnight is the overnight shift (22:00 to 06:00), and you tick "ends next day". - **People** are assigned to a shift. Assigning the same person twice is the same as once. - **Exceptions** are per date: a **day off** takes someone out of the shift, and an **extra** puts someone in who is not on it. Saving the same exception again corrects it instead of stacking a second truth.
"How many people" is what produces the coverage alert. Leave it at zero and that shift never reports a gap: zero means "we are not measuring this one", not "nobody needed".
Coverage: the question before all the others
On the supervision wallboard, **Coverage** crosses the three sources and answers *is anyone on?*
- **Covered** — the shift asks for two, two are receiving work. - **Short** — someone on shift is on a break. It is a warning, not a fire. - **Uncovered** — the queue is open, the shift exists, and nobody on it is receiving. Red. - **Off shift** — no shift is running right now. The panel says "we do not know", never "0 of 0". If the queue is open anyway, it warns you that there is no plan for this hour.
It also lists the **gaps ahead** for the next 24 hours: windows where fewer people are scheduled than the shift asks for. That is the one you fix before it becomes a red square.
Occupancy and adherence
With planned time in place, two numbers become honest, in the Analytics tab of the service console:
- **Occupancy** = time handling work divided by time connected. It comes from the presence ledger, which each person feeds by changing their own state. This is not a time clock and has no employment-law meaning. - **Adherence** = time connected inside the shift divided by the shift itself. It is the supervisor's question: was the plan met?
Where the server does not know, the screen shows a dash. No roster means no adherence, and writing 0% there would be an unfair conversation with a real person.
Redistributing from the wallboard
The supervision board used to only watch. **Rebalance** proposes a plan, move by move, with the reason written next to each one:
- unassigned and at risk of missing its deadline
- unassigned
- the owner is unavailable (offline, on a long break, out of rotation)
- the owner is over their ceiling
A case whose owner is present and within their ceiling is **not** moved. Taking work away from someone who is working would make the customer start the conversation over with a different person.
You read the plan first, then press the button. The work is handed over through the same path the console uses, so permissions, history, triggers and closing guards all still apply, and each case comes back with what happened to it. Every run is recorded: that is the answer to "who sent this to me?".