Section 01

Two channels, three threads at a time, every hour of the year

Chat is roughly two thirds of the volume and email a third. The two behave differently enough that they are staffed, measured and ramped separately.

	Live chat	Email and ticket
Share of volume	~65%	~35%
Concurrency	2.5–3.5 conversations at once	Sequential, worked from a queue
What the customer expects	An answer inside 30 seconds	A considered answer inside 4 hours
Where it goes wrong	Speed at the cost of accuracy	A queue that ages quietly overnight
Who works it	Chat-primary agents	Email and escalation agents
Ramp position	Introduced in week three, one chat at a time	Where every new agent starts

New agents start on email, not chat. Email is forgiving of a slow answer and unforgiving of a wrong one, which is exactly the right way round for someone still learning the product. Chat is added one conversation at a time from week three.

Contact mix

What customers actually write in about

Indicative of the current placement. The mix matters because it sets how much product depth an agent needs before they are useful, and therefore how long the ramp is.

Contact type	Share	Typical resolution	Depth required
How-to and configuration	~35%	Answered from the knowledge base with a cited article	Moderate — the largest single category
Account, access and login	~24%	Resolved in the conversation against the client's admin tooling	Low, but identity checks are strict
Troubleshooting a defect	~20%	Diagnosed, reproduced where possible, escalated with evidence	High — this is where escalation accuracy is won or lost
Billing and subscription	~15%	Explained; anything that moves money is routed to the client	Moderate; the boundary is absolute
Feature requests and roadmap	~6%	Logged, acknowledged honestly, never promised	Low, but judgement is required

Never promise a roadmap item. An agent who tells a customer a feature is coming has created a commitment the client did not make. The approved response acknowledges the request, logs it, and says plainly that it cannot confirm timing.

Productivity basis

The numbers the rate was built on

Read concurrency as a range and not a target. An agent pushed to 3.5 chats will hit the number and lose the quality score, and the composite is weighted so that trade never pays.

Measure	Standard after ramp	Notes
Productive hours per agent per shift	7.0 of an 8.0-hour shift	Excludes breaks, briefings, coaching and training
Concurrent chats	2.5–3.5	Three is the working target; above that quality falls before speed does
Chats handled per shift	45–65	Varies with contact type and time of day
Emails handled per shift	40–55	Higher on account and billing queues, lower on troubleshooting
Occupancy	80–85%	Deliberately not higher; overnight bands run quieter by design
Schedule adherence	≥ 95%	The measure that protects coverage
First-contact resolution	≥ 75%	Influenceable — depends on product and permissions as well as skill
Knowledge-base citation rate	≥ 90%	An answer that cannot be sourced is an answer no one can audit

Indicative monthly volume at 30 seats

At 22 working days and 7.0 productive hours per agent, a 30-seat pod delivers approximately 4,620 productive hours a month. On the current channel mix that is roughly:

Productive hours: 4,620
Chats handled: ~21,000
Emails handled: ~9,500

Volume is not contracted. What is contracted is the seat and the coverage. Contact volume moves with the client's own customer base, its release cycle and its marketing — none of which the floor controls. A quiet month is not a shortfall and a spike is not a breach; both are managed through the roster.

Ramp

The concurrency curve

Six weeks, and it cannot be rushed. Billing starts on the agent's first live conversation; training time is never billable.

Period	What the agent handles	Expected output
Weeks 1–2	Email only, reviewed before sending	50% of the email standard
Week 3	Email plus a single chat at a time	60% of the combined standard
Week 4	Two concurrent chats	75%
Week 5	Two to three concurrent chats	85%
Week 6 onward	Full concurrency, all queues	Full standard

Pushing concurrency early is the classic failure on this process. An agent moved to three chats in week three will produce a quality score that takes a quarter to recover, and the composite is weighted so that quality carries more than speed — which means the shortcut costs more than it saves.

Next step

The conversation is the easy part. The roster is not.

The coverage page sets out the arithmetic, the four bands, and the one credit mechanism on this process that has real teeth.

Read the coverage arithmetic See the scope of work