E-commerce

How to build a support knowledge base for your e-commerce team

June 28, 2026

“Where is my order?” Three agents, three phrasing variants, three different return policies. The customer reopens the ticket. CSAT drops. This is not a motivation issue: it is the lack of a centralized, versioned, and maintained support answer base.

Ad hoc customer service scripts (#templates) speed up sending; they do not replace an internal source of truth. Ecommerce Circle observes that a library of 25 well-maintained macros can reduce response time by 60 to 80% (Ecommerce Circle, macro library 2026).

This guide #102 covers REP structure, taxonomy, ticket mining, governance, and bot sync. Distinct angle: internal ops and response quality, prerequisites before bot training (#103).

Summary

How does a knowledge base differ from customer service scripts and macros?

What structure should be given to each response sheet?

How to organize taxonomy by categories?

What editorial standards should be imposed on the team?

Which tool should be hosted for a Shopify store?

How to build the base from your tickets?

How do we maintain and version the sheets without obsolescence?

How can we get agents to adopt the database?

How to synchronize the internal database, bot, and help center?

How to manage multilingual and variations by segment?

How does Qstomy leverage your knowledge base?

Which operational playbooks should be launched this week?

How does a knowledge base differ from customer service scripts and macros?

Clarifying the roles of the e-commerce support response database prevents three divergent truths in your helpdesk.

Operational Comparison

Internal Database REP-*: source of truth, complete policy, escalations, exceptions.
Gorgias Macros: delivery shortcuts derived from the database, Shopify variables.
Customer Service Scripts: occasional situational phrasing, no centralized governance.
Public Help Center: simplified customer version, self-service.
Guidance Bot: validated excerpt, not raw macros since 2026.

Gorgias specifies that macros no longer serve as a knowledge source for the AI Agent: useful content must migrate to help center articles or guidance entries (Gorgias, update AI Agent 2026). The REP database thus becomes the pivot between human agents, macros, and the bot.

Distinct from customer service support templates and helpdesk vs chatbot vs KB. See also support strategy (#25).

Why Invest Now

Target ticket reopen rate: under 5%; above 10%, a systemic quality signal (ScreenMeet, help desk metrics 2026). Inconsistent responses = reopened tickets, chargebacks, slow agent onboarding. From 3 agents upwards, a Notion or Guru database is no longer a luxury.

What structure should be given to each response sheet?

Each entry in the customer service response library follows an identical template so that any agent can find the same information within ten seconds.

Required fields

ID: REP-SHIP-001 (category + sequential number).
Agent title: "WISMO order in transit" (not the customer's question).
Customer formulations: "parcel not arrived", "tracking stuck", "where is my order".
Short response: chat, 2-4 sentences max.
Long response: email, context + next step.
Variables: {{order_number}}, {{tracking_url}}, {{customer_first_name}}.
Linked Policy: link to the official Shopify return page.
Escalation: supervisor criteria (VIP, amount > €300, public complaint).
Gorgias Tags: wismo, tier1, shipping.
Owner + last review: name, date, draft / published / deprecated status.

Concrete example REP-RET-003

Question: "35 days after delivery, can I return?" Short response: 30-day policy + portal link. Long: manager exception if < 45 days and loyal customer. Escalation: VIP or LTV > €500. Decision tree at the top of the card if/then before the customer text. SLA reference: SLA grid (#101).

CaseKit Rule

Macro + personalization 5-10 s before sending: the agent reads the ticket, chooses the card, adjusts a contextual line (CaseKit, macros vs AI 2026).

How to organize taxonomy by categories?

An e-commerce support taxonomy aligned with your Gorgias tags avoids duplicates and blind spots.

Build prioritization

Export tags 90 days: top 20% of categories = 80% of volume. Start with WISMO, returns, delivery time, declined payment. Hero products: dedicated REP-PRE sheets for specific SKU questions. Convention ID: REP-[CAT]-[NUM] with CAT = SHIP, RET, PAY, PRE, ACC, B2B, SUB, FRA.

See tagging conversations, products generating tickets, compatibility questions. Seasonal sheets REP-BFCM-* archived in January: BFCM preparation.

What editorial standards should be imposed on the team?

Consistent support response standards protect the brand and reduce reopenings.

DTC writing rules

Structure: empathetic acknowledgment → factual response → concrete action → next step.
Chat length: 4 sentences max; email: 150 words excluding signature.
Clear policy: "Return within 30 days in brand new condition" instead of "in accordance with our policy".
Tested variables: never leave empty {{placeholder}} in production.
Forbidden: promises outside policy, customer blaming, emojis if premium brand.

Good vs Bad

Bad: "No returns accepted." Good: "Our return window is 30 days post-delivery. Here is the portal: [link]." Mandatory personalization: first name, product, order number if available in Shopify sidebar.

Review before publication

Policy up to date? Variables tested on a real ticket? Escalation documented? Peer review by a second agent? Sensitive topics (allergy, health claim): immediately escalate using form REP-ESC-001. See complaints and loyalty (#94), measure response quality.

Which tool should be hosted for a Shopify store?

The choice of the support knowledge base tool depends on the team size and the existing stack.

Options by Stage

Notion: MVP 1-8 agents, searchable wiki, draft/published permissions.
Gorgias Help Center noindex: Internal agent KB, meta robots noindex (Gorgias, internal KB).
Guru: cards + browser extension, Slack, 10+ agents.
Confluence: enterprise versioning, 20+ multi-brand agents.

Decision Criteria

Search < 15 s (mystery agent test). Gorgias ticket extension for macro insert. Version history. Export API for bot sync. 48-hour MVP: Notion + 20 top ticket sheets + 20 macros linked by REP ID.

Gorgias Macros: bundle actions (reply + tags + status), not text only (eesel AI, 2026 macros). Naming: `[SHIP] WISMO transit` rather than "Response 3". See knowledge base structure, Shopify integration.

How to build the base from your tickets?

The support ticket mining transforms 500 resolved conversations into actionable files.

Six-step process

Gorgias 90-day export: closed tickets, tags, CSAT.
Cluster by tag or intent (top 30 groups).
Select best agent response: CSAT 5/5, no reopen.
Internal tag `gold_reply` on exemplary tickets.
Normalize into REP template (section 2).
Gap analysis: clusters without a good response → draft new ones.

Signals to exploit

CSAT 1-2 tickets: missing or outdated file. Bot unresolved export: new candidate intents. Hero SKU launch: 3 REP-PRE files before ads go live. Pre-purchase objection mining for conversion section.

AI draft from gold ticket: human editing mandatory before publishing. See conversation analytics, detect objections, conversations → PDP.

How do we maintain and version the sheets without obsolescence?

A knowledge base without maintenance becomes inaccurate within 90 days after a promo, carrier change, or return policy update.

Operational schedule

Weekly: review 5 top-volume sheets.
Monthly: 30-minute macro audit (Ecommerce Circle): archive < 5 uses/30 days.
Quarterly: 2 hr gap workshop, merge duplicates, deprecated cleanup.
Post-launch: dedicated sheets within 48 hrs (SKU, BFCM promo).

Policy change workflow

Policy update (e.g.: return 30 → 45 days).
Owner lists impacted REP-RET-* sheets.
Draft v2, head of support review.
Publish base → update macros → guidance bot → public help center.
Slack #support-kb-updates: 3-line changelog.

Deprecated status with redirect to new sheet, never sudden deletion. An obsolete article costs more than no article: reopen rate and disputes. Promo codes: sync REP-PROMO at the same time as email marketing. See support promo codes.

How can we get agents to adopt the database?

Answer-based governance transforms a passive wiki into a daily team tool.

Clear Roles

Knowledge manager: review calendar, templates, adoption metrics.
Category owners: writing REP-SHIP, REP-RET, etc.
QA lead: sample 5% tickets vs. article used.
Agents: suggest edit button, flag outdated.

Onboarding D+1 to D+5

Day 1: read top 10 high-volume articles. Day 2: shadow + linked macros. Day 3: 10-scenario quiz (80% required before solo queue). Week 2: QA feedback on 5 tickets. Part-time 2 days/week: searchable database = immediate productivity.

Adoption Metrics

Target macro usage rate > 70% Tier 1. Custom reply rate too high = insufficient database. High macro edit rate on an article = update the template with agent improvements. Target FCR 70%+ reduces reopen rate below 5% (Alexander Jarvis, reopen rate). See DTC support playbook, automation mistakes.

How to synchronize the internal database, bot, and help center?

The bot help center answer base sync prevents the customer from reading one policy while the agent applies another.

Content Pyramid

Top: master REP internal base, exceptions, and escalations.
Middle: Gorgias macros + bot guidance (100 entries max Gorgias 2026).
Bottom: public help center, simplified paragraph extracted from REP.

Single Publish Workflow

Published REP sheet → export customer-safe paragraph → sync tier 0 bot intent → Gorgias macro with same ID. A single publish gate (head of support) avoids triple editing. Authorized differences: shorter bot, internal plus exceptions. Core policy identical everywhere.

Before bot training: 50 published sheets, zero drafts in production. Macros containing useful info: convert to guidance, do not leave as a standalone macro (Gorgias update 2026). PDP accordion = simplified REP-PRE extract.

See clean data before bot (#103), train Shopify chatbot, FAQ reduce tickets, self-service (#28).

How to manage multilingual and variations by segment?

Expanding the multilingual response base without version chaos requires unique ID discipline.

Master + translations approach

Master FR: REP-SHIP-001 source.
Language variants: REP-SHIP-001-EN, same policy content.
Market policy: fork REP-RET-003-EU vs REP-RET-003-UK if deadlines differ.
Deprecated sync: identical status for all languages.

Segment variants

`vip_variant` field: documented commercial gesture. B2B: referral to account manager. Marketplace: platform rules. Dropshipping delay: subset REP-DROP. Digital: REP-DIGI access and license.

International export: top 30 sheets translated as a priority. See international support (#100), VIP escalation, dropshipping (#96), digital support (#95), chargebacks.

How does Qstomy leverage your knowledge base?

Qstomy consumes your validated response base to align bot and agents on the same policy, connected to Shopify.

Key capabilities

KB Import: Notion, Gorgias help center, CSV answer sheets.
Intent mapping: customer phrasing → sheet ID.
Channel variants: short chat, long email if escalated.
Shopify variables: order, tracking, first name injected.
Gap report: questions without sheets → answer backlog.
Webhook publish: sheet update → guidance re-sync.

Quantified DTC Scenario

Cosmetics brand 120 orders/day, 4 agents, 85 tickets/day, macro usage 34%, reopen rate 14%, CSAT 3.8.

Migration: 65 answer sheets from ticket mining, Qstomy sync + 40 ID-linked macros. 10-week result: macro usage 34% → 76%, reopen rate 14% → 4.2%, median handle time −38%, bot tier 0 covers 48 intents out of 65 sheets, CSAT 3.8 → 4.3. Support Head: 6 hrs/month maintenance vs 18 hrs rework before base.

Explore AI support, reduce AI tickets, chatbot KPIs, request a demo.

Which operational playbooks should be launched this week?

Playbook 1: REP template + 10 top tags cheat sheets

Create the section 2 template in Notion. Export Gorgias top 10 tags for the last 90 days. Draft REP-SHIP-001 WISMO, REP-RET-001 standard return, REP-PAY-001 declined payment as a priority.

Playbook 2: macro ↔ cheat sheet linkage

For each published cheat sheet, create a Gorgias macro `[CAT] Title` with Shopify variables. Test: mystery agent resolves a WISMO ticket in < 60 s using only the macro + 1 personalized line.

Playbook 3: gold_reply mining

Tag 20 CSAT 5/5 tickets. Extract 5 new cheat sheets this week. Compare agent wording vs cheat sheet: gap = template update.

Playbook 4: guidance bot migration

List macros containing policy (return, delay, refund). Convert them into Gorgias guidance entries or noindex help center articles. Check that the AI Agent no longer quotes any obsolete old macros.

Playbook 5: 30-min monthly review

Archive macros with < 5 uses/30 days. Update the 3 most edited cheat sheets by agents before sending. Publish Slack changelog. Measure macro usage rate and reopen rate the following week.