Most support queues have a priority field. Far fewer have a shared definition of what the values mean. The result is predictable: one agent marks a password reset "High" because the customer sounded upset, another marks a broken checkout "Normal" because it came in through a low-priority channel. Your SLA reports then measure mood, not urgency.
A triage rubric fixes that. Here's how to build one that two people, in two time zones, will score the same way.
Step 1: Score two dimensions, not one
"Priority" mixes two separate questions. Ask them separately:
Impact: how many people or how much money is affected?
| Score | Definition |
|---|---|
| 3 | All customers, a whole tenant, or anything touching payments, security or data loss |
| 2 | One customer's team, or one feature broken for several customers |
| 1 | One user, with a workaround available |
Urgency: what happens if this waits until tomorrow?
| Score | Definition |
|---|---|
| 3 | Work is stopped now, or a deadline lands within 4 hours |
| 2 | Work is slowed; a deadline lands within 2 business days |
| 1 | Inconvenient, no deadline |
Every definition is something an agent can check from the ticket, not a feeling. "Customer is angry" isn't in either table on purpose.
Step 2: Map the scores to a priority
Multiply or use a lookup grid. A grid is easier to read at 2 a.m.:
| Urgency 1 | Urgency 2 | Urgency 3 | |
|---|---|---|---|
| Impact 3 | P2 | P1 | P1 |
| Impact 2 | P3 | P2 | P1 |
| Impact 1 | P4 | P3 | P2 |
Then attach a response target and an owner to each level, for example:
- P1: acknowledge in 15 minutes, page the on-call engineer, update the customer every hour.
- P2: acknowledge in 1 hour, assign to a named agent, escalate to tier 2 if not resolved in 4 hours.
- P3: first response same business day.
- P4: first response within 2 business days; candidate for a help-center article.
Use your own numbers. The point is that each priority triggers a specific action, not just a colour.
Step 3: Add override rules, and keep the list short
Some tickets should skip the grid. Write these down explicitly so nobody has to judge it on the spot:
- Any suspected security incident or data exposure → P1, regardless of scores.
- Any report that a payment was taken twice or not refunded → at least P2.
- A ticket reopened for the second time → raise one level and assign to a lead.
Three to five overrides is plenty. If the list grows past that, your impact and urgency definitions need work.
Step 4: Write five worked examples
Definitions alone leave room for interpretation. Real examples close it. Pick tickets from your own history and score them in the rubric document:
- "I can't log in, reset link isn't arriving." One user, workaround (support can resend or verify manually), no deadline → Impact 1, Urgency 2 → P3.
- "Our whole team gets a 500 error on the dashboard." One customer's team, work stopped → Impact 2, Urgency 3 → P1.
- "Export to CSV puts dates in the wrong format." Several customers, workaround exists, no deadline → Impact 2, Urgency 1 → P3.
- "Payroll export fails and payroll runs at 3 p.m. today." One team, deadline within 4 hours → Impact 2, Urgency 3 → P1.
- "Can you add dark mode?" Feature request → route out of the incident queue entirely.
Example 4 is the one that teaches the most: the same feature as a routine bug becomes P1 because of the deadline. That's what the urgency axis is for.
Step 5: Calibrate before you roll it out
Pull 20 recent tickets. Have three agents score them independently with the rubric, without discussing. Then compare:
- Where all three agree, the definitions are working.
- Where they disagree, find the word that caused it ("several", "slowed", "deadline") and tighten it.
- Repeat with a fresh 20 until disagreements are rare and explainable.
This takes an hour and saves weeks of arguing about SLA breaches.
Step 6: Put the rubric where the decision happens
A rubric in a wiki nobody opens doesn't exist. Practical placements:
- Make Impact and Urgency required dropdowns on the ticket form, and compute priority from them with a helpdesk automation.
- Pin the grid and the overrides as the first macro or saved reply agents see.
- Include the five worked examples in onboarding for every new agent, including contractors and offshore team members, who often join without the hallway context everyone else picked up.
Step 7: Review it monthly
Once a month, pull every ticket where the priority was changed after triage. Each change is either an agent error (coaching) or a rubric gap (edit the definitions). Track the count; it should fall.
A one-page template
TRIAGE RUBRIC v1 (owner: ____, reviewed: ____)
Impact: 3 = all customers / payments / security / data loss
2 = one team, or one feature for several customers
1 = one user, workaround exists
Urgency: 3 = stopped now, or deadline < 4 h
2 = slowed, or deadline < 2 business days
1 = no deadline
Grid: see table; P1–P4
Overrides: security → P1; double charge → ≥P2; 2nd reopen → +1 level
Targets: P1 15 min / P2 1 h / P3 same day / P4 2 days
Examples: (paste your five)
A consistent rubric is what makes a distributed support team work: when everyone scores tickets the same way, a hand-off between shifts doesn't change what "urgent" means. We set up this kind of structure when we build dedicated support teams, and it's the first document we ask for.
Written by the TrueNext Global team. Drafted with AI writing tools for TrueNext Global.
Top comments (0)