Live chat staffing calculator for your busiest hour.

Work out how many agents you need on chat to answer in time, how many to schedule once breaks and meetings are counted, and how that changes when a chatbot resolves part of the volume.

Made by the team behind Conecto, which sells website chat software. How we keep it useful

Demand
per hour

New chats that start in your peak hour.

minutes

From first reply to close, measured at your usual concurrency.

chats

How many conversations one person handles in parallel.

Service target
%

sec

%

The busiest you want agents to be, on average.

People and automation
%

Time paid but not on chat: breaks, meetings, training, leave.

%

Chats that end before reaching a person. Use 0 for live chat only.

Prefilled values are illustrative assumptions, not benchmarks.

Agents on chat at peak

3

97.3% answered within 60 seconds

Agents to schedule

5

After 30% shrinkage

Average wait

4 seconds

Occupancy 44.4%

With a chatbot resolving 25%

2

on chat at peak, 3 to schedule

A planning estimate for one hour. It assumes steady arrivals and nobody leaving the queue.

Service level by number of agents

Agents on chatAnswered within 60 sAverage waitOccupancy
10%Queue grows without limit100%
272.9%68 seconds66.7%
397.3%4 seconds44.4%
499.8%Under 1 second33.3%
5100%Under 1 second26.7%
6100%Under 1 second22.2%

The highlighted row is the smallest team that meets both your service target and your occupancy ceiling.

A chatbot that answers documented questions first can take the peak off your queue.

How the calculation works

The calculator uses Erlang C, the standard queueing formula behind most contact-center staffing tools. It estimates the chance that a new chat has to wait, given how much work arrives and how many people are available.

Chat is different from phone because one agent can handle several conversations at once. The calculator approximates this by treating each agent as a number of parallel slots equal to their concurrency, with the handle time spread across those slots.

Workload in slots = chats per hour × handle time in seconds ÷ 3,600 ÷ concurrency

Chance of waiting = Erlang C for (agents × concurrency) slots and that workload

Answered in time = 1 − chance of waiting × e−(slots − workload) × target seconds ÷ (handle time ÷ concurrency)

Average wait = chance of waiting × (handle time ÷ concurrency) ÷ (slots − workload)

Occupancy = workload ÷ slots

Agents to schedule = agents on chat ÷ (1 − shrinkage), rounded up

The calculator increases the number of agents until the share answered in time meets your target and occupancy is at or below your ceiling.

Worked example

With 40 chats in the busiest hour, 8 minutes of handle time and 2 chats per agent, the workload is 2.67 slots. 3 agents on chat answer 97.3% of chats within 60 seconds, with an average wait of 4 seconds and 44.4% occupancy. With 30% shrinkage, that means scheduling 5 people for that hour.

If a chatbot fully resolves 25% of chats, the busiest hour needs 2 agents on chat and 3 scheduled. The chatbot only helps with questions it can answer from documented information, so check what share of your chats are like that before relying on it.

What the model leaves out

  • People leaving the queue. Erlang C assumes everyone waits. In reality some visitors give up, which makes the formula slightly cautious.
  • Concurrency slowing chats down. Handling several chats at once makes each one take longer. Measure handle time at your usual concurrency so the estimate reflects it.
  • Uneven arrivals. A campaign or outage can bring chats in bursts. Plan the busiest hour you actually see, not the daily average.
  • Chats a chatbot starts but can’t finish. A handoff still reaches a person, sometimes with a head start. Only fully resolved chats reduce the workload here.

How to find your numbers

Chats in your busiest hour
Export chat start times from your chat tool or help desk and count them per hour over a few weeks. Use a typical busy hour, not a one-off spike.
Handle time
Measure from the agent’s first reply to the chat closing, including wrap-up notes.
Concurrency
Start with two. Three is manageable for simple questions; one suits complex or sensitive ones.
Service target
Decide how quickly visitors should get a first reply from a person. On a website, people are often waiting on the page, so the target is usually measured in seconds rather than minutes.
Shrinkage
Add up breaks, meetings, training, time off and other work as a share of paid hours.

Formula and guidance last reviewed on September 23, 2026. Your inputs stay in your browser: nothing is stored or sent.