Playbook · Zendesk

Zendesk ticket categorization best practices

How to map the C1-C2 methodology to Zendesk's fields, tags, and nested dropdowns so your reports finally answer the question "why are people contacting us?".

1. Why most Zendesk categorizations fail

Nine out of ten Zendesk instances we audit share the same problem: a single "Category" dropdown that grew to 40+ values, half of which mean the same thing ("Refund", "Money back", "Restitution"). Agents pick whatever is closest, managers can't trust the reports, and the AI project everyone wants to start has no clean training data to learn from.

The fix is not "add more fields". The fix is a two-layer structure with strict rules about what belongs where — that's what C1-C2 is.

2. The C1-C2 model in 60 seconds

C1 is the topic — the object of the conversation ("Order", "Return", "Product"). C2 is the reason — what the customer actually wants ("Where is it?", "Damaged on arrival", "How does it work?"). Every ticket gets exactly one C1 and one C2.

• C1 is a short, closed list (usually 8-15 values). It rarely changes.
• C2 depends on C1. "Where is it?" only makes sense under "Order" or "Return", not under "Product".
• Together, C1×C2 gives you 60-120 meaningful combinations — enough resolution for routing, staffing and AI, without the 400-row nightmare.

If you want to build your own C1-C2 tree first, use our free C1-C2 Builder — it ships with templates for e-commerce, repair, SaaS and more.

3. Mapping C1-C2 to Zendesk fields

Zendesk gives you three primitives to work with: ticket fields, tags, and the built-in About field. Here is the mapping we recommend on every implementation:

4. Nested dropdowns: setup that scales

Zendesk's nested dropdown uses :: as the separator inside a single ticket field. That's how you enforce that a C2 can only appear under its parent C1.

Rules that keep the tree clean

• One owner per branch. If nobody owns "Product", nobody prunes it — and it will grow to 40 leaves in six months.
• Never more than 12 C2s per C1. If you need more, the C1 is probably two topics glued together (e.g. split "Account" into "Login" and "Billing").
• No "Other" as a C2. It becomes the default landing spot and kills your data. Force a real choice or add the missing branch.
• Retire values, don't delete them. Historic tickets keep the label; new tickets can't select it. Zendesk supports this with the "Available for agents" toggle.

5. Tags vs. custom fields: use both, but differently

Custom fields are for structure you report on. Tags are for signals that don't need a full field — usually agent-added context or trigger-added flags.

• Use fields for: C1, C2, channel, brand, product line, resolution outcome.
• Use tags for: vip_customer, ai_deflected, escalated_to_l2, seasonal campaigns.
• Never use tags for C1/C2. Tags are free-form, sorefund, refunds and Refund all coexist — destroying the reports you built C1-C2 to enable.

6. Triggers, macros and routing

Once C1-C2 is a real field, you can build triggers that would previously have needed fragile keyword matching:

• Route by C1. "Return" tickets go to the returns queue; "Product" tickets go to the pre-sales group. One trigger, no keyword regex.
• Auto-assign SLAs by C2. "Order::Where is it?" is high volume, low complexity, tight SLA. "Product::Warranty claim" is the opposite.
• Gate macros on C2. A refund macro should only be selectable when C2 is a refund-related value — reduces mis-sends and legal risk.
• Feed AI on clean labels. Zendesk's Advanced AI (and every third party auto-tagger) trains far better on 80 clean C1×C2 combinations than on 400 messy tags.

7. Reporting in Explore without the noise

The whole point of C1-C2 is that one Explore report answers the leadership question "what are people contacting us about?". Build these three dashboards first:

• Volume heatmap (C1 × C2). Instantly shows which cells drive contact — usually 5-8 combinations account for 60% of volume.
• Handling time by C2. Reveals which reasons are eating agent time and are worth automating or self-servicing.
• Deflection potential. Combine C2 volume with "solved on first reply" and channel — that's your AI/self-service backlog, prioritized.

Because C1-C2 is a single nested field, every Explore attribute is one dropdown click — no calculated attributes, no tag-parsing, no maintenance.

8. Rollout checklist

1. Design your C1-C2 tree in the Builder before touching Zendesk admin.
2. Calibrate the tree with 3-5 agents on 50 real tickets. Aim for >85% agreement.
3. Create the nested dropdown field, hide the old category field (don't delete — historical reports still need it).
4. Update triggers and SLA policies to read from the new field.
5. Retrain agents in 30-minute blocks. Include a "when in doubt" decision rule per C1.
6. Review the C1 × C2 heatmap monthly. Prune dead cells, split overloaded ones.