Operations analytics

The analytics dashboard turns your live operation into numbers you can act on. Every metric is derived from the same records your team works, so the report and the reality never drift apart. Agents can view the dashboard; exporting the underlying data is reserved for Admin and Owner.

Queue volume and waits

Volume is how many live chats entered the queue in the window. Alongside it, Hark reports wait times as percentiles, not just an average, because averages hide the tail. The median (p50) tells you the typical wait; the p90 tells you what your slowest tenth of customers experienced, which is usually where complaints come from.

Abandonment and conversion

• Abandonment counts the chats a customer gave up on before an agent picked them up. The abandonment rate is that count against total volume.
• Conversion counts the chats that reached a live agent or were resolved, against volume. Together these tell you whether your queue is fast enough to hold customers.

Average handle time

Average handle time (AHT) is the live "talking" time: how long an agent is actively engaged in a chat. It is reported as an average and as percentiles, and can be broken down per agent so you can see spread across the team rather than one blended number.

Utilization and capacity

Because presence is recorded over time, Hark can report utilization: how much of their available time agents spent actually handling chats. Paired with volume by hour, this feeds capacity and staffing views that show when you are over- or under-covered, so scheduling is driven by data rather than guesswork.

Transfers

Transfer counts show how often chats move between agents and teams. A high transfer rate into one team can reveal a routing gap or a training need.

SLA compliance

The SLA section reports compliance per timer (first response, next response, resolution, queue wait): what share of conversations met each target. Breaches are broken down so you can see where they cluster, for example by team, inbox, or priority, and fix the cause rather than chase symptoms.

Reading the numbers honestly

Every metric here is grounded in real events: queue entries, presence intervals, the activity trail, and resolution outcomes. Where a rate cannot be computed (for example, no volume in the window) Hark shows no value rather than a misleading zero.