customer-support-ops

Files (7)

• README.md
• SKILL.md
• evals.json
• escalation-matrix.md · references/
• macro-templates.md · references/
• sla-playbook.md · references/
• triage-automation.md · references/

SKILL.md (first lines)

## Key principles

1. **First response time is king** - The metric customers feel most is how quickly
   someone acknowledges their problem. A fast FRT buys goodwill and time. Every SLA
   must prioritize FRT above all others. Measure it first; everything else is secondary.

2. **Triage before solving** - An untriaged queue is a random queue. Agents working in
   random order guarantee high-priority problems wait behind low-priority ones. Triage
   assigns priority, tier, and routing - it does not solve the problem.

3. **Macros save hours, not minutes** - One macro used 50 times a day saves 50x the
   time invested in writing it. Expand the library whenever agents write the same reply
   more than once a week. Review quarterly. A bad macro is worse than no macro.

4. **Escalation paths must be clear** - Ambiguous escalation is the leading cause of
   tickets stalling. Every agent must know, without asking, exactly when and where to
   escalate. If an agent has to think about whether to escalate, the path is not clear.

5. **Measure everything** - Support intuition degrades under volume. Track FRT,
   resolution time, CSAT, first contact resolution rate, escalation rate, and queue age.
   Review weekly. Data surfaces problems before customers do.

---

## Core concepts

### Support tiers

| Tier | Typical customers | Support entitlement |
|---|---|---|
| Community / Free | Trial, free plan | Docs and community forum only |
| Standard | Paying customers | Email, SLA 24h FRT |
| Professional | Growth plan | Email + chat, SLA 8h FRT |
| Enterprise | Large contracts | All channels, SLA 1h FRT, named contacts |

**Tier assignment rule:** Tier is set at account creation from the billing plan and
must trigger automatic re-routing in the ticketing system. Never rely on agents to
manually route by tier - automate it.

### SLA components

| Component | Definition | Measured from |
|---|---|---|
| First Response Time (FRT) | Time from ticket creation to first agent reply | Ticket creation |
| Next Response Time (NRT) | Time between agent replies after customer responds | Customer reply |
| Resolution Time (RT) | Time from ticket creation to closed status | Ticket creation |

**SLA clock rules:**
- FRT clock starts immediately on ticket creation, including nights and weekends,
  unless the tier contract specifies business hours only.
- Clock pauses when a ticket enters "Pending Customer" status (waiting on customer).
- Clock never pauses for internal notes or transfers between agents.

### Ticket lifecycle

```
Submitted -> Triaged -> Assigned -> In Progress -> Pending Customer
                                        |                  |
                                    Escalated          Reopened
                                        |
                                    Resolved -> Closed (auto after 7 days)
```

Each transition must have a defined owner and time constraint. Tickets sitting in
"Triaged" for more than 30 minutes indicate a routing or staffing problem. Tickets
in "In Progress" beyond the resolution SLA need a manager flag.

### Queue management

**Queue states to monitor:**

| State | Definition | Action threshold |
|---|---|---|
| New | Submitted, not yet triaged | > 15 min: trigger triage alert |
| Breached | Past FRT SLA | Escalate to lead immediately |
| At-risk | Within 20% of SLA window | Flag for prioritization |
| Aging | Open > 5 days with no update | Manager review required |
| Stalled | No agent activity > 24 hours | Auto-assign to queue lead |

---

## Common tasks

### Design a triage system

**Priority matrix (impact x urgency):**

| | Low urgency | High urgency |
|---|---|---|
| **High impact** | P2 - schedule soon | P1 - respond now |
| **Low impact** | P4 - backlog | P3 - respond today |

**Priority definitions:**

```
P1 - Critical:  Service down, data loss, security issue, or revenue blocked.
                FRT target: 1 hour. Assign immediately. Page on-call if needed.

P2 - High:      Core feature broken, workaround difficult, or VIP affected.
                FRT target: 4 hours. Pull from queue before P3/P4.

P3 - Normal:    Feature degraded, workaround exists, standard customer.
                FRT target: per tier SLA (8h or 24h).

P4 - Low:       Cosmetic issue, how-to question, feature request.
                FRT target: 48 hours. May be batch-processed.
```

**Triage checklist (run on every new ticket):**
1. Assign priority using the matrix above.
2. Confirm customer tier from CRM. Upgrade priority if enterprise.
3. Check for duplicate tickets from the same account - merge if found.
4. Apply tags: product area, issue type, channel source.
5. Route to the correct queue or agent group.
6. Set SLA clock based on tier and priority.

### Set up SLAs

**SLA matrix by tier and priority:**

From mkurman/zorai, MIT.

More from mkurman/zorai