E-commerce
June 28, 2026
Without a taxonomy, your helpdesk counts tickets but does not tell you why customers are writing. "Return" or "delivery" hide actionable root causes: missing size guide, forgotten tracking, poorly applied promo.
Pattern Owl recommends a three-level hierarchy with 30 to 50 active themes (Pattern Owl, feedback taxonomy 2026). Gorgias advises fewer than 30 well-defined core tags (eesel AI, Gorgias tags guide 2026).
This guide #135 explains how to create a comprehensive support ticket taxonomy: dimensions, naming conventions, helpdesk implementation, governance, and dashboards. It complements tagging conversations (#117) with the overall structure, not just the pain points method.
Summary
Why use a support ticket taxonomy in e-commerce?
An e-commerce support ticket taxonomy is the common language that transforms ticket volume into product, ops, and content decisions.
What a taxonomy enables
Prioritize: which friction points to address first
Automate: reliable helpdesk rules
Measure: CSAT and handling time per intent
Train the bot: intents aligned with human tags
Align teams: same definition for "size return"
Without a taxonomy
Every agent invents their own tags. "parcel", "shipping", and "wismo" are triple-counted. Dashboards lie. The bot and humans speak different dialects. From 300 tagged tickets/month, trends by category become clear.
Concrete example
A fashion brand sees 18% of "return" tickets without a sub-tag. After taxonomy v2 (size_return, quality_return, refund_delay_return), the product team discovers that size_return accounts for 62% of returns on the jeans collection and triggers a PDP measurement guide brief. Without a sub-tag, the team would have generalized "returns are up" without any targeted action.
How does it differ from conversational tagging?
The article tagger conversations (#117) teaches the pain points method and the agent workflow. #135 sets up the complete taxonomy architecture.
Scope #135
Dimensions: intent, funnel, channel, product, resolution
Nomenclature: naming conventions, official glossary
Governance: who creates, merges, archives a tag
Lifecycle: v1 launch, v2 quarterly, annual audit
Dashboards: helpdesk views by dimension
Also completes segment by funnel and conversational analytics.
Which dimensions should be included in the taxonomy?
A mature taxonomy combines multiple dimensions, not just a single flat tree.
Dimension 1: intent (mandatory)
Main reason: wismo, return, pre_purchase_product, payment, order, account. Mandatory main tag upon closure.
Dimension 2: funnel
pre_purchase, post_purchase, post_delivery. Crosses intent to prioritize conversion vs. after-sales service.
Dimension 3: contact channel
email, chat, instagram, whatsapp, tiktok, amazon. Distinct from the sales channel.
Dimension 4: transverse
sentiment: neutral, frustrated, chargeback_threat
resolved_1st_contact: yes/no
escalation: senior_agent, ops, legal
sku_collection: optional if product volume
Maximum 3 intent tags + 2 transverse tags per ticket. Beyond that: analytical noise.
Useful cross-matrix
Cross intent × funnel × channel to identify leaks. Example: pre_purchase_product × instagram × pre_purchase = size questions in DM before purchase. wismo × email × post_delivery = standard parcel tracking. return_exchange × chat × post_delivery = real-time after-sales support friction. This matrix feeds weekly support priorities and content briefs by channel.
How should levels 1, 2, and 3 be structured?
Recommended 3-level hierarchy structure for DTC Shopify.
Level 1: domains (10-15 max)
wismo, pre_purchase_product, order, return_exchange, payment, delivery, product_support, account, marketplace, spam.
Level 2: friction sub-intents
Wismo example: missing_tracking, blocked_scan, delay_exceeded_promise, package_not_received. Return example: incorrect_size, missing_guide, refund_delay, return_label.
Level 3: optional context
collection, carrier, marketplace_amazon, promo_black_friday. Use only for targeted action (e.g., return + jeans_collection + 40% volume).
Tag creation thresholds
New sub-tag if 20+ unclassifiable tickets/month. Merge if < 5 tickets/month over 90 days. Fewer than 20 active themes = loss of nuance; more than 60 = overlaps.
Example wismo tree
wismo (n1) → missing_tracking (n2) → carrier_colissimo (n3 optional). wismo (n1) → delay_exceeded_promise (n2) → promo_delivery_48h (n3). wismo (n1) → package_not_received (n2) → signature_required (n3). The agent always chooses n1 + n2; n3 only if requested by the macro or ops escalation.
How do I name and document each tag?
The tags support nomenclature avoids case-sensitive duplicates and drift.
Conventions
Format: lowercase, underscores or hyphens, never spaces
Specific: payment-failed, not payments
Client language: taille_petit, not ecart_fit_negatif
Dimension prefix: funnel_preachat, channel_instagram
Notion tag sheet (1 page glossary)
For each tag: 2-line definition, 3 verbatim examples, 1 counter-example, owner action (product, ops, content), associated macro, automation rule if applicable.
Before adding a tag: "What action do we trigger with this data?" (MESA, Shopify 2026 tags).
How do I implement taxonomy in Gorgias or Zendesk?
Implement the helpdesk taxonomy in 6 technical steps.
Create level 1 and 2 tags in Settings (Gorgias) or custom fields (Zendesk)
Configure mandatory fields upon closing: main tag + funnel
Create Views by family: open wismo, open return, VIP priority
Pre-tagged Macros: WISMO macro applies wismo + tracking_communique
Rules WHEN/IF/THEN: intent shipping + tracking exists → auto-reply + tag
Export test of 100 tickets: verify tag distribution
Sync Shopify tags
Customer tag vip, order tag gift → Gorgias routing rules (US Tech Automations, routing tags 2026). Support taxonomy ≠ Shopify tags, but map vip → support escalation.
Common implementation mistakes
Creating 80 tags at once without training agents: adoption < 40% in the first week. Mandatory tags without an associated macro: agents choose the vaguest tag to close quickly. Forgetting Views by family: the taxonomy exists but no one filters by it. Best practice: launch with 15 core tags, train for 1 hour, measure inter-agent agreement, then expand.
How can you automate tagging without compromising quality?
Automation of ticket tagging: hybrid AI + rules + human.
Phase 1: keyword rules
Inexpensive, rigid. Refine AND/OR: "return" + "label", not "return" alone. Gorgias Rules: WHEN new ticket IF message contains tracking THEN tag wismo.
Phase 2: AI helpdesk intent
Gorgias intent detection, Zendesk Intelligent Triage. AI suggests tag, agent validates for 90 days. Then auto-tag if confidence > 90% on stable categories (simple wismo).
Phase 3: pre-tag bot
Escalated ticket inherits bot tags. Agent corrects if poorly detected. Log corrections = training. See choosing questions to automate, helpdesk vs chatbot.
How to govern and evolve the taxonomy?
Taxonomy Governance: roles, ritual, changelog.
RACI
Taxonomy Owner: Head of Support (creation, merging, archiving)
Contributors: senior agents suggest sub-tags
Validators: product/ops if the tag impacts their scope
Rituals
Weekly 20 min: confusing tags reported by agents. Monthly: inter-agent agreement on a sample of 30 tickets, target 85 %+. Quarterly: audit of tags with < 5 uses, merging duplicates (cancellation vs cancel-order). Slack changelog for each new sub-tag.
Drift signals
"Miscellaneous" or "other" tag rises to top 5 = obsolete taxonomy. New collection without product sub-tag = spike in unclassified tickets.
Process for requesting a new tag
Slack form #support-taxonomy: agent proposes tag, 3 verbatims, estimated volume, expected action. Owner validates or merges within 48 hours. Reasoned refusal if duplicate or insufficient volume. Prevents the proliferation of "one-shot" tags created after an isolated ticket but never reused.
How do you link taxonomy to dashboards and KPIs?
The taxonomy powers actionable support dashboards.
KPIs by tag
Volume and MoM growth by sub-tag
FRT and resolution time by intent
CSAT by tag family
48-hour repeat contact: incomplete response
Bot deflection by intent aligned with taxonomy
Actionable readings
Volume x revenue impact matrix. Slack alert if promise_overdue_time +50% vs week N-1. Cross-reference tag + SKU + BFCM season. See response quality, support SLA, products generating tickets.
How do you connect taxonomy, bots, macros, and content?
Taxonomy is the bridge support → product → content → bot.
Loops by top volume tag
Tag guide_taille_absent +20 tickets/month → brief PDP guide
Tag promo_non_appliquee → sync macro + checkout UX
Tag tracking_manquant → fulfillment ops alert
Tag intent bot unmatched → chunk corpus bot
Each Gorgias macro maps 1 primary tag. Bot intents = same names as level 1 tags. See product insights from support, merchandising conversations, knowledge base.
Monthly loop template
Each month, for the #1 volume tag: (1) volume and CSAT, (2) root cause in 1 sentence, (3) action owner (product, ops, content, bot), (4) review date. Document in Notion. In 6 months, you have a measurable history of pain point reduction, not just a declining ticket counter.
How does Qstomy power the ticket taxonomy?
Qstomy aligns bot and human ticket taxonomy with actionable data.
Features
Intent mapping: bot intents = official tags
Monthly tag report: volume, growth, unmatched
Sub-tag suggestion: unmatched cluster to brief taxonomy
Handoff tags: transcript + bot intents transferred to agent
Export glossary: sync Notion taxonomy
Quantified DTC Scenario
DTC Brand 2,800 tickets/month, taxonomy v1 42 sub-tags. Before Qstomy: 23% of tickets tagged "miscellaneous". After bot alignment + 8-week unmatched report: "miscellaneous" 6%, 12 new data-driven sub-tags created. Bot deflection +19 points on WISMO and returns. 3 blog articles published from top tags (#127 workflow). Taxonomy lead audit time: -4 hours/month.
Explore AI support, Shopify, request a demo.
Which operational playbooks should be launched this week?
Playbook 1: Taxonomy Workshop v1 (1 day)
Export 200 tickets from the last 30 days. Verbatim clustering in the morning. Validate 10-15 domains + 3 sub-tags for the top 5 families in the afternoon. Agreement test on 30 tickets, target 85%+. Detailed method: tagging conversations (#117). Timeframe: 1 day.
Playbook 2: Notion Glossary 1 page
Table: tag, definition, example, owner, macro, rule. Share with agents on D+1. 10 fictional tickets onboarding quiz.
Playbook 3: Gorgias Implementation
Create tags, Views, 5 starter Rules (WISMO, VIP, return, spam, negative Instagram). Mandatory tag upon closing. Timeframe: 3 hours.
Playbook 4: Weekly Tags Dashboard
Google Sheet: top 10 tags by volume, % growth, CSAT, 48-hour repeat. Review on Friday for 20 mins with support + product. 1 priority action per week.
Playbook 5: Quarterly Audit
Merge tags with < 5 uses. Archive completed campaign tags. Recalibrate definitions for top 5 tags. Goal: miscellaneous < 10% of volume.
Useful Linking
A living taxonomy is better than a perfect, frozen taxonomy. Start simple, measure, and iterate quarterly.
Enzo
June 28, 2026
