Two time zones, one shared inbox: routing a 19-agent support team
Meridian Ledger is a B2B SaaS company with 19 support agents across Boston and Dublin. It uses HubSpot Service Hub Professional with one shared Support inbox, one four-stage ticket pipeline, and three channels: email, live chat, and phone. Email arrives in the shared inbox and agents self-assign. Chat creates a ticket only about 40% of the time, and phone calls are logged manually after the fact. At 3,100 tickets a month, 31% of tickets get their first substantive reply from the wrong specialist team, 22% miss the six-hour first-response target, and 8% of conversations receive duplicate replies because two agents open the same thread. The support director has approval to add four agents and wants to do so next quarter. The CFO asks whether the team is understaffed or misconfigured. Service Hub is already paid for; the team has not used routing rules, team inbox permissions, or required ticket properties.
1. Before approving four new agents, what would you audit, and what would you change first? Explain the reasoning.
Read the worked answer
Start by treating the four-agent request as a hypothesis, not a conclusion. The waste signals are stronger than the volume signal: 8% duplicate replies means two agents are doing the same work, and 31% wrong-team first replies means work is being rerouted after it has already started. Self-assignment in one shared inbox creates cherry-picking, invisible ownership, and no skill matching. Chat that creates a ticket only 40% of the time and manually logged phone calls mean the queue is also undercounted, so the team may be both misconfigured and carrying hidden demand.
First 30 days:
- Make every email, chat, and phone contact create or link to a ticket; require topic, priority, language, and segment before closure.
- Replace self-assignment with team-based routing: route by topic and language to a team, then round robin or load-balance within the team.
- Restrict the shared inbox so agents work from ticket views, not from an unassigned pool.
- Add an exception/triage queue for unmatched tickets and a named owner for it.
- Fix duplicate handling with a conversation owner and collision detection where available.
Only after these changes would I compare contacts per agent and handling time against demand. If backlog age and SLA breaches persist after duplicate work and misroutes are removed, then add headcount.
2. Design the routing and ownership model for email, chat, and phone. Include what should be routed, who owns what, and how exceptions are handled.
Read the worked answer
Use a tiered model. All channels create a ticket. Email and chat route first by topic (billing, onboarding, API, account access) and language; high-value or P1 accounts can route to a named pod. Phone is logged as a ticket with call outcome and recording link, not typed in later. The ticket owner is accountable for resolution; the conversation assignee is whoever is currently handling the thread, and the two should not be confused.
Tier 1 handles known issues, account access, and triage; Tier 2 handles product defects and integrations. Chat should route to available Tier 1 agents with a fallback to email if no one is available. After-hours uses a separate queue and an on-call escalation for P1. Unmatched tickets go to a triage queue with a 30-minute ownership target. Exceptions: named-account requests go to the account pod; security or data-loss reports bypass round robin and page the on-call. The routing rules should write the assigned team and reason to the ticket so reporting can show misroutes.
3. Which metrics would you present after 60 days to prove the changes worked, and what would make you conclude you still need headcount?
Read the worked answer
Present a before/after operating dashboard, not just ticket volume. Baseline: first-response SLA attainment, median and 90th-percentile first response, reassignment rate, duplicate-contact rate, backlog age, chat-to-ticket creation rate, phone contacts logged, CSAT, and reopened rate. After 60 days, the changes should reduce reassignments and duplicates and raise SLA attainment without reducing contacts handled per agent. If contacts per agent is already near capacity and backlog age is still rising after waste is removed, that is the headcount case. If handling time is stable but demand is spiking, use temporary coverage or deflection before permanent roles. Do not use average first response alone; the 90th percentile shows the enterprise tail.