E-commerce
June 28, 2026
The ticket is closed. Twenty minutes later, a "Rate your experience 1 to 5" email arrives. Response rate: 9%. The unhappy customer has already posted on Google. The satisfied customer doesn't click. You are steering a CSAT without knowing why it is dropping.
Perspective AI estimates that static web surveys peak at an 8-15% response rate, compared to 20-35% for an in-app conversational exchange that digs into the "why" behind the rating (Perspective AI, CSAT 2026). Decagon reminds us that collecting feedback without acting on it erodes trust more than not asking at all (Decagon, best practices 2026).
This guide #239 covers post-support feedback collection via chatbot. Distinct from general feedback content and NPS timing (#53): here, we focus on the CSAT/CES flow right after bot or human resolution, directly within the chat stream.
Summary
Why collect feedback in the post-resolution chat?
The post-support chatbot feedback captures satisfaction while the emotion is still fresh, without email friction.
Limits of email surveys
Delay, promo inbox, cold form. The customer no longer connects the rating to the agent or bot that helped them. Impossible to tag intent, thread duration, bot vs human.
Three in-context chat benefits
Response rate: 2 to 4× vs email according to conversational benchmarks
Rich context: ticket_id, intent, agent, duration attached to the score
Adaptive probe: 5/5 vs 2/5 do not receive the same follow-up question
DTC fashion example
Brand with 2,100 resolutions/month, email CSAT 11% response. Post-close in-chat bot flow: 34% response, qualitative drivers on 78% of ratings. Time-to-action on detractors −62%.
How does it differ from NPS, response quality, and chatbot KPIs?
Four measurement contents, one moment: just after support resolution.
NPS timing (#53)
NPS (#53): when to send global NPS. The #239: micro CSAT/CES post-ticket in the widget.
Quality responses (#116)
Quality (#116): QA accuracy/tone audits. The #239: voice of the customer on the same interaction.
Chatbot KPI (#11)
Chatbot KPI (#11): deflection, FCR, ROI. The #239 feeds the CSAT bot vs blended segment component.
Product insights (#33)
Insights (#33): mining support verbatims → product. The #239 structures the CSAT collection upstream.
Promise #239
Triggers, adaptive flow, detractor escalation, payload analytics, KPI, playbooks.
When should the post-support feedback flow be triggered?
The chatbot feedback trigger must follow a confirmed resolution, not an abandonment.
Five valid triggers
Bot FCR: customer confirms "Resolved" or closes without reopening for 5 mins
Closed human handoff: agent marks solved + customer inactive for 3 mins
Async email ticket: status closed + "1 quick question" link widget
Portal self-serve: return validated, tracking sent
Recovery after anger: P1 resolution, 2-hour delay before CSAT (to avoid raw emotion)
Do not trigger if
Ticket reopened within 24 hours, P0 escalation in progress, customer leaves mid-handoff, or conversation < 2 turns (likely bounce). Denser recommends post-support as a priority trigger (Denser, feedback chatbot 2026).
Anti-spam frequency
Max 1 post-support CSAT / customer / 7 days. Exclude if CSAT has already been answered on parent ticket.
Which metrics to collect: CSAT, CES, and verbatims?
The post-support feedback payload balances numerical scores with qualitative depth.
CSAT (Priority 1)
"How would you rate this interaction?" 1-5 scale or emojis. Decagon: industry benchmark 75-85%, best-in-class > 85% (Decagon).
Optional CES (effort)
"How easy was it to get a resolution?" 1-7. Relevant post-handoff or multi-turn pathways.
Verbatim driver (Priority 2)
Open-ended question adapted to the score. AI Perspective: moving from "CSAT 78%" to "driver #1 detractors = agent delay" (Perspective AI, why behind score).
Automated Metadata
ticket_id, intent, channel, bot_only Y/N, agent_id, handle_time, goodwill_gesture Y/N, LTV tier. Feeds into the support dashboard.
What adaptive conversational flow depending on the rating?
The conversational CSAT flow adapts the following steps to the score, rather than using a fixed form.
Branch 4-5 (promoter)
Score 4-5
"What helped you the most?" (choices: speed, clarity, agent, resolution)
Option: Trustpilot review link if score is 5 and consent is given
Thank you + closing
Branch 3 (passive)
"What could we improve?" 1 choice + optional free text field. No auto-escalation.
Branch 1-2 (detractor)
Empathy: "We are sorry."
"Was the issue mainly related to: delay / response / resolution / other?"
Free verbatim
Proposal: reopen ticket or supervisor callback within 4 hours
Target duration
60-90 seconds, max 4 steps. Quick buttons + "Skip" option without penalizing null CSAT.
How to route negative feedback without losing the customer?
A post-support CSAT detractor is a second chance, not an archive.
Auto-escalation rules
Score 1-2: ticket reopened with tag
csat_recovery, assign supervisorKeywords: scam, lawyer, chargeback → P0 even if score not entered
VIP + score ≤ 3: Slack alert + tier 2 queue
Bot-only + score ≤ 2: intent + corpus review, do not blame customer
Recovery macro CSAT-REC-01
"Thank you for your honest feedback on ticket #[X]. A support manager will contact you back within 4 business hours to fix what went wrong. Would you like a callback?"
Closed loop
Decagon: measure if CSAT insight triggers action within 72 hours. Otherwise, do not solicit the customer again (feedback fatigue). See angry customers (#214).
How to connect feedback, helpdesk, and bot improvement?
The feedback ops pipeline transforms scores into actions, not a monthly slide.
Minimum webhook payload
{csat_score, ces_score, driver_tag, verbatim, ticket_id, intent, bot_only, agent_id, timestamp}
Helpdesk sync
Ticket tag csat_[1-5], custom score field + verbatim. Gorgias/Zendesk: "detractors 7 d" view for agent coaching.
Bot loop
Top bot-only detracting drivers → corpus gap or intent. Aligned with bot anti-hallucination. Rule of thumb: fix source documentation, not just the model.
20-min weekly ritual
Last 10 detractors: root cause ops vs. content vs. customer expectation. 1 concrete action per week (macro, REP sheet, bot fix).
Which anti-patterns should be avoided in post-support collection?
Five post-support feedback errors kill response rates and trust.
CSAT before resolution
"Are you satisfied?" when the package is still lost. Permanent negative biased score.
Long form
8 questions = abandonment. Max 2 questions + 1 verbatim.
Collecting without acting
Detractor ignored for 5 days → customer will never respond again.
Agent CSAT linked to bonus without context
Gaming scores, refusal of difficult cases. Weight by ticket complexity.
Mixing support and product NPS
"Would you recommend the brand?" at the end of customer service confuses interaction and brand. Keep NPS for a separate trigger (#53).
How to test the feedback flow before production?
The QA flow CSAT bot validates triggers, branches, and escalations.
Minimum of 20 scenarios
5 bot FCR → trigger OK, branch 5 and branch 2
5 human handoff → agent_id metadata
3 ticket reopen < 24 h → no CSAT
3 detractor → ticket reopened + alert
2 skip « Skip » → clean closure without bugs
2 anti-spam 7 d → no double CSAT
Shadow mode
Proposed flow without auto-escalation for 1 week. Measure completion rate and verbatim length.
Light A/B
Emojis vs 1-5 numerical on 500 resolutions. Keep the format with the highest completion rate.
Which KPIs should be used to manage post-support feedback?
Measure the post-support CSAT program, not just the average score.
Collection KPIs
Response rate: target > 25% in-chat
Flow completion rate: score + at least 1 follow-up
% actionable verbatim: > 15 characters, driver tag
Quality KPIs
Average CSAT bot / human / blended segment
Top 3 detractor drivers per month
Recovery rate: detractor contacted → CSAT ≥ 4 after recovery
Detractor time-to-action: target < 4 h
Dashboard
Columns: score, intent, bot_only, agent, driver, recovery Y/N. Cross-reference with Chatbot KPIs (#11).
How does Qstomy collect post-support feedback?
Qstomy triggers the CSAT flow upon closing with the full Shopify ticket context.
Feedback Features
Post-resolution trigger FCR bot or closed handoff
Adaptive flow 1-2 / 3 / 4-5 branches
Analytics payload intent, agent, duration, gesture
Detractor escalation reopen + supervisor alert
Anti-spam 1 CSAT / 7 days / customer
Export drivers to weekly bot + agents review
Quantified DTC Scenario
Beauty brand, email CSAT 10% response rate, unknown drivers. In-chat post-close flow + detractor escalation. After 8 weeks: 31% response rate, blended CSAT 4.2/5 (+0.3), detractor recovery 58%, 4 bot intents corrected via top driver "unclear response".
Explore Shopify integration, AI customer support, request a demo.
Which operational playbooks should be launched this week?
Playbook 1: triggers (3 h)
List 5 triggers from section 3, exclusions, 7-day anti-spam rule.
Playbook 2: 3-branch flow (4 h)
Draft steps from section 5, quick buttons, Skip option, max duration 90 s.
Playbook 3: detractor escalation (2 h)
Section 6 rules, CSAT-REC-01 macro, Slack alert, 4 h SLA.
Playbook 4: webhook + dashboard (1 d)
Section 7 payload, helpdesk tags, 7-day detractors view.
Playbook 5: weekly ritual (20 min)
10 detractors, 1 corrective action, Notion log.
Useful links
Post-support feedback is not a substitute for operational excellence: it makes it visible. When every resolution ends with a rating, a reason, and a loop on dissatisfied customers, CSAT ceases to be an isolated number and becomes the driver of improvement for your customer support and your bot.
Enzo
June 28, 2026
