Concurrency (Chat Concurrency)
The number of live conversations a single agent handles at the same time, mostly used for chat and messaging channels.
An agent handles multiple live conversations and you need to find where extra parallel work begins reducing quality.
The team plots response time and CSAT against open chats at each reply, then sets the cap below the observed knee.
Using one fixed concurrency multiplier for every agent, issue type, and customer state.
Concurrency = concurrent chats / agents online (or, from times: customer handle time / agent working time)Concurrency is what makes chat cheaper per contact than voice: a phone agent handles one call at a time, but a chat agent can juggle several conversations while customers type, read, and wait. It is the lever operators reach for when they want to lift throughput without adding headcount, and it is the number a workforce-management plan for messaging channels is really built around.
The simplest measure is a real-time ratio — concurrent chats divided by agents online — so 15 active chats across 5 agents is a concurrency of 3. It can also be derived from times: divide the customer-facing handle time (including the gaps where the customer is reading or typing) by the agent's actual working time on that chat. The two views answer different questions: one is load right now, the other is how much genuine idle an agent had to fill.
What it hides: concurrency is an average that flattens the worst moments. A stated concurrency of 3 does not mean three calm parallel conversations — it means that during peaks the agent is switching between all three at once, and every added chat pushes response times and quality down while nudging escalation and canned-reply rates up. It also quietly inflates handle-time and occupancy math: a chat that looks fast on paper may just be a customer left waiting between an agent's replies. Chase a higher concurrency target and you can hit it on the dashboard while the experience each customer actually receives gets slower and thinner.