E-commerce
July 1, 2026
"Your chatbot recommended a men's product to me, but I am a woman." "The suggestion makes no sense for my sensitive skin." "Fix your recommendations or I will stop buying." Three messages where a customer unhappy with a widget, email, or AI assistant suggestion opens a ticket, whereas a feedback+correction bot could recognize, recollect, and propose grounded catalog alternatives.
An e-commerce AI chatbot bad recommendations does not replace IRECO-FLOW agents (#439). It reads IRECO-MAP, welcomes structured feedback, checks purchase history, proposes a grounded alternative, records the correction for the merchandising loop, and routes offensive P1 cases to #439.
This guide #440 covers bot_ireco_* intents, IRECO-BOT flow, and ireco_bot KPIs. Distinct from bad reco customer service (#439) and contextual reco setup: here, we focus on the AI use case for feedback and recommendation correction to preserve trust.
Summary
Why does a feedback recommendation bot reduce trust-related support tickets?
An avoidable bad recommendation ticket is born when the customer cannot find how to report an absurd suggestion and obtain a credible alternative before writing to support or unsubscribed.
Five triggers for bot-resolvable bad recommendation tickets
Bad bot suggestion: assistant suggested incompatible SKU
Widget wrong: off-topic PDP carousel reported in chat
Already owned: customer mentions repromoted purchased product
Correction request: "what do you really recommend?"
Structured feedback: filling out a form via conversational bot
McKinsey points out that 71% of consumers react negatively to "fake personalization" (McKinsey, personalization 2026). Gorgias Shopping Assistant documents recommendations based on cart browsing and dialogue (Gorgias, assistant 2026). Tidio estimates that a Tier 1 recommendation feedback bot reduces support tickets caused by bad recommendations by 46-58% vs. static FAQ (Tidio, chat stats 2026).
Angle #440 vs adjacent content
Bad recommendation customer service #439: IRECO-FLOW merch flag opt-out execute. #440 = tier 1 bot ack feedback correction alternative.
Contextual recommendations: conversion engine setup. #440 = post-failure customer complaint.
Assistant vs recommendation #17: strategic choice. #440 = Al quality correction loop.
Product questions: PDP usage compatibility. #440 = "your suggestion was bad".
Governance #142: saying vs doing. #440 applies to bot recommendation feedback.
DTC Example
Skincare 180 bad recommendation tickets/year baseline #439. After IRECO-BOT: ireco_bot_resolution 84%, ireco_bot_ticket_delta -51%, ireco_correction_accept_rate 62% customer accepts bot alternative.
Trust recovery
A bot that admits its mistake + offers 1-2 grounded SKUs beats a bot that defends its initial suggestion. IRECO-BOT enforces "no defend".
Widget PDP thumbs down baseline
Deploy T1 before peak season. Capture bad recommendation feedback pre-ticket when recommendation widgets traffic increases 3x for Black Friday.
How does IRECO-BOT differ from IRECO-FLOW #439 and contextual recommendations?
Bot feedback reco, playbook agents ireco, and setup engine reco: three complementary layers.
Matrix bot ireco → role
#440 IRECO-BOT: ack feedback, collect reason, verify owned, suggest correction grounded, log learning
#439 IRECO-FLOW: merch flag execute, opt-out Klaviyo, P1 gesture, social DM
Contextual reco: widget rules engine setup merchandising
Product questions bot: pre-purchase compatibility without reco complaint
Router upstream
“Bad recommendation”, “suggestion makes no sense”, “already purchased product suggested again”, “correct your advice”, “feedback reco” → IRECO-BOT. “Offensive reco allergy”, “refund gesture”, “complex unsubscribe” → handoff #439 IF-6. “Which product to choose?” without complaint → product questions bot.
Data IRECO-BOT
IRECO-MAP JSON #439, order history 90 days, cart session, bot transcript prior suggestion, Shopify catalog API, feedback form fields, exclude purchased list.
Promise #440
Policy IRECO-BOT, 12 intents bot_ireco_*, flow IB-1 to IB-8, guardrails no defend grounded re-reco, KPI ireco_bot_*.
Learning loop
bot_ireco_feedback_log → weekly merch + bot corpus review. Distinct real-time suggestion engine retrain ops.
Nosto Rebuy feedback loop
Third-party reco engines may accept exclude SKU API. Document IRECO-MAP engine_vendor handoff merch for rule push.
Which bot_ireco_* intents should be configured?
Twelve bad recommendations bot intents cover feedback, correction, and routing.
Twelve bot_ireco intents
bot_ireco_bad_suggestion : bot/widget/email suggestion complaint
bot_ireco_ack_apologize : empathetic error acknowledgment
bot_ireco_feedback_collect : structured wrong_category owned duplicate
bot_ireco_verify_owned : order API SKU already purchased
bot_ireco_suggest_correction : 1-2 SKU grounded alternative
bot_ireco_explain_why_shown : honest widget email bot source
bot_ireco_opt_out_reco_email : segment reco unsub trigger
bot_ireco_preferences_update : zero-party skin type gender size
bot_ireco_feedback_form_link : /pages/reco-feedback deep link
bot_ireco_log_learning : session export IF-8 corpus
bot_ireco_handoff_439 : P1 offensive allergy merch complex gesture
bot_ireco_route_product_q : product question without reco complaint
Session tags
ireco_bot, ireco_bot_feedback, ireco_bot_corrected, ireco_bot_resolved, ireco_bot_handoff_439. Distinct product_bot, ireco_agent.
Triggers T1-T5
T1 : post-bot suggestion thumbs down "Not relevant". T2 : keyword bad recommendation/absurd suggestion. T3 : widget footer feedback opens bot. T4 : /pages/reco-feedback bot embed. T5 : proactive after ireco email click unsubscribe hesitates chat.
Intent priority routing
Thumbs down T1 auto-routes bot_ireco_bad_suggestion before generic product intents. Prevents wrong macro misfire.
How to apply the IRECO-BOT flow in eight steps?
The IRECO-BOT flow guides grounded feedback reco IRECO-MAP #439 and catalog.
Eight steps IB-1 to IB-8
IB-1 Greeting: "I have taken your feedback on our product suggestions into account."
IB-2 Auth context: email optional, session cart, prior bot transcript
IB-3 Classifier intent: bot_ireco_* section 3
IB-4 Match IRECO-MAP: source engine opt_out correction rules
IB-5 Read data: orders 90 days, SKU complained, catalog alternatives
IB-6 Respond: ack + feedback collect + correction or explain
IB-7 CTA: add-to-cart alternative | feedback form | opt-out reco
IB-8 Close/log: ireco_feedback_log handoff #439 if P1
IB-6 bad_suggestion flow
Step 1 bot_ireco_ack_apologize. Step 2 bot_ireco_feedback_collect reason menu. Step 3 IB-5 verify_owned if cited. Step 4 bot_ireco_suggest_correction max 2 SKU grounded with why 1 sentence each.
IB-5 suggest_correction rules
Exclude complained SKU, purchased 90 days, OOS, wrong gender collection tags. Read metafields compatibility. No LLM invent SKU.
IB-6 explain_why_shown
Honest cite IRECO-SOURCE: "suggestion based on similar cart" not "perfect profile". bot_ireco_explain_why_shown template.
IB-8 learning log
JSON: session_id, sku_bad, reason_code, sku_suggested, accepted_y/n. Weekly export merch + bot training review.
IB-2 thumbs down T1
Post suggestion UI thumbs down → auto IB-1 without client typing complaint. Highest ireco_bot_resolution rate.
IB-8 JSONL schema
Fields: ts, session_id, sku_bad, reason_code, sku_suggested[], accepted, ireco_source, customer_id_hash. Merch BI ingest.
Which IRECO-BOT policy should be documented?
The IRECO-BOT bad recommendations policy governs ack, correction, and learning.
Eight IRECO-BOT rules
Acknowledge never defend: no "the suggestion was relevant"
IRECO-MAP grounded: source engine opt_out from JSON #439
Correction catalog API only: bot_ireco_suggest_correction IB-5 rules
Max 2 alternatives: no 8 SKU carousel post-complaint
Verify owned before re-suggest: exclude purchased 90 d
Log every feedback: bot_ireco_log_learning IB-8 mandatory
P1 handoff #439: offensive allergy action bot cannot grant
Max 6 turns feedback flow: then form CTA or agent
RAG source corpus
/pages/reco-feedback, IRECO-MAP #439, IRECO-SUP #439, collection compatibility metafields, exclude rules doc.
Zero-party preferences
bot_ireco_preferences_update: store skin_type gender size session profile for future suggestions. Consent quotes CNIL marketing profile.
Monthly bot reco quality review
Audit 20 transcripts: defended suggestion count zero, correction grounded, timely P1 handoff.
Thumbs down widget integration
PDP reco carousel thumbs down feeds bot_ireco_bad_suggestion T1 with placement_id context IB-5.
CNIL preferences storage
bot_ireco_preferences_update session profile: cite retention and opt-out in privacy policy link.
Which guardrails protect trust and catalog?
The authorized vs prohibited ireco bot matrix protects trust and product accuracy.
Authorized bot actions
Ack apologize bot_ireco_ack_apologize
Collect feedback reason structured menu
Suggest 1-2 SKU grounded IB-5 exclude rules
Explain source widget email bot honest
Trigger reco email opt-out if requested
Update session preferences zero-party
Log IB-8 learning export
Handoff #439 P1 offensive allergy action
Prohibited bot actions
Defend original bad suggestion
Invent SKU correction hallucination
Re-suggest complained SKU same session
Promise merch fix timeline bot
Grant action discount bot unless IRECO-MAP pre-approved catalog
Blame client navigation history
Anti-hallucination correction
Template: "Here is [SKU name] [price]: [1 sentence why]" from catalog JSON only. See hallucination prevention.
Allergy skincare guard
If feedback cites allergen conflict → immediate bot_ireco_handoff_439 P1. Bot does not suggest alternative without metafield verify.
No upsell post apology
Correction SKUs same price tier or lower unless client asks upgrade. Trust recovery not AOV spike.
Reason code merch mapping
wrong_category → collection rules audit. already_owned → exclude engine fix. price_tone_deaf → price band rules.
How to connect IRECO-MAP #439 and Shopify catalog?
The bot feedback reco integration combines IRECO-MAP, orders API, and catalog lookup.
IRECO-BOT read fields
prior_suggestion_sku : bot/widget/email SKU complained
ireco_source : widget_pdp cart email bot
orders_90d_skus[] : exclude purchased
cart_line_skus[] : exclude duplicate
catalog_alternatives : same collection compatible tags
customer_preferences : zero-party profile if set
Shopify Storefront API product recommendations + custom exclude purchased filter (Shopify, recommendations 2026).
IB-5 correction algorithm lite
Input complained SKU + reason_code. Filter: in stock, not owned 90 d, match preferences, same need category. Rank by rating margin policy. Return top 2.
Bot transcript attach
If ireco_source bot: read prior assistant message SKU cited. IB-8 log links bad suggestion turn index.
Qstomy Integration
See Shopify integration + IRECO-MAP RAG sync on merch rule update.
Feedback form webhook
bot_ireco_feedback_form_link pre-fill sku reason from IB-6 session. Reduce duplicate data entry.
Storefront MCP lookup_catalog
IB-5 correction uses lookup_catalog grounded variants not LLM browse. Sync with train chatbot Shopify corpus.
What triggers thumbs down, widget, and UX for IRECO-BOT?
The IRECO-BOT UX deployment maximizes feedback capture and correction accept rate.
Five widget placements
Thumbs down reco T1: post suggestion bot or carousel
Widget footer T3: "Suggestion not relevant?" chat chip
/pages/reco-feedback T4: bot embed guided form
Post-email click T5: hesitated unsub opens bot opt-out vs correct
Assistant after bad pick: auto IB-1 on thumbs down
Thumbs down on bot suggestion
Shopping assistant message footer 👎 → IB-1 immediate ack. Captures ireco before ticket. Target 70% resolution without agent.
Structured feedback menu IB-6
Buttons: Wrong category | Already purchased | Not my style | Too expensive | Other. Maps reason_code log merch.
Correction accept tracking
Click add-to-cart on bot_ireco_suggest_correction SKU → ireco_correction_accept_rate KPI.
A/B ack copy
IRECO-ACK variant A vs B: measure ireco_bot_csat and correction accept 4 weeks.
Cross-sell cart widget feedback
Cart drawer reco block thumbs down → bot_ireco_bad_suggestion with cart context IB-5 exclude cart SKUs.
Email reco unsub vs correct choice
T5 bot offers fork: opt-out reco segment OR update preferences bot_ireco_preferences_update before unsub all.
Which KPIs does ireco bot measure?
The bad recommendations bot KPIs link deflection, correction, and quality loop.
Eight key metrics
ireco_bot_resolution_rate: resolved without handoff / ireco_bot sessions
ireco_bot_ticket_delta: decrease in tickets #439 vs baseline
ireco_feedback_capture_rate: structured feedback / thumbs down T1
ireco_correction_accept_rate: add-to-cart correction / bot suggestions
ireco_bot_handoff_439_rate: P1 routed agents
ireco_repeat_complainer_bot: 2+ ireco_bot same email 90 days post-bot
ireco_bot_csat: satisfaction tag ireco_bot resolved
ireco_learning_log_volume: IB-8 exports / month merch reviewed
DTC Benchmark
ireco_bot_resolution 82-90 %, ireco_bot_ticket_delta -45-55 %, correction_accept 55-68 %, ireco_bot_csat > 4.2/5, repeat_complainer < 6 % post bot.
Monthly Dashboard
reason_code breakdown, top SKU complained, correction SKU accept, source widget vs bot vs email, learning log → merch fixes closed loop.
Defend count audit
Transcript scan « suggestion was relevant » count zero target. Rule 1 violation = P2 incident.
NPS recovery cohort
NPS ireco_bot resolved vs ireco agent only. Target bot +8 pts vs agent baseline.
Thumbs down to resolution funnel
T1 click → IB-6 complete → correction accept or opt-out. Drop-off step analysis monthly UX.
Which anti-patterns should be avoided on bot feedback reco?
Ten bad bot recommendation anti-patterns to ban.
1. Defend bad suggestion
Rule 1 forbidden. CSAT collapse and social screenshot risk.
2. Invent correction SKU
Rule 3 catalog API. Hallucination product = chargeback if wrong item.
3. Re-suggest same SKU complained
Exclude complained SKU session permanently IB-5.
4. Skip IB-8 log
Rule 6 learning loop broken. Same error repeats.
5. Upsell premium post apology
Trust recovery not AOV. Same tier or lower correction default.
6. Blame client browsing
IRECO-SUP #439 rule 3 applies bot tone.
7. 8 SKU carousel correction
Rule 4 max 2. Overwhelms frustrated client.
8. Bot geste unauthorized
Rule 7 handoff #439 geste_policy. Bot promo promise forbidden.
9. No thumbs down T1
Missing lowest friction feedback capture. Tickets instead.
10. Confondre product question
bot_ireco_route_product_q if no reco complaint. Different flow.
11. Allergy suggest without verify
Handoff P1. Metafield ingredients required before alternative skincare.
12. Merch log ignored
ireco_learning_log_volume without merch review = theater not improvement.
13. Correction without why sentence
IB-6 must include 1 phrase why per SKU. Naked link feels random second bad reco.
How does Qstomy correct, learn, and preserve trust?
Qstomy on Shopify: IRECO-BOT ack feedback, catalog grounded correction, order exclude purchased, thumbs down T1 trigger, learning log IB-8 export, handoff #439 P1 pre-filled fields transcript.
ireco bot Qstomy Capabilities
ireco_ack_flow: IB-6 apologize structured
ireco_feedback_menu: reason_code collect
ireco_correction_engine: IB-5 top 2 SKU
ireco_owned_exclude: orders 90 d filter
ireco_learning_export: IB-8 weekly JSONL
ireco_handoff_439: P1 fields attach
Pipeline #440 → #439
Bot tier 1 feedback correction log. Agents P1 complex merch gesture. Shared IRECO-MAP learning corpus.
Quantified DTC Scenario
180 ireco tickets/year baseline #439.
After IRECO-BOT Qstomy: ireco_bot_resolution 86%, ireco_bot_ticket_delta -49%, ireco_correction_accept_rate 64%, ireco_bot_csat 4.3/5.
Explore customer support and request a demo.
Corpus sync training
See train Shopify chatbot for IRECO learning log → RAG monthly.
Sales assistant quality
See sales assistant for suggestion quality loop with ireco_bot feedback.
Weekly learning review ritual
30 min support + merch + bot owner: top 10 sku_bad from IB-8, assign rule fix owner, close loop in Notion.
What is the checklist for deploying IRECO-BOT?
IRECO-BOT Checklist (12 steps)
Validate IRECO-SUP #439 + IRECO-MAP /pages/reco-feedback
Export IRECO-MAP + exclude purchased rules JSON
Configure 12 intents bot_ireco_* section 3
Implement flow IB-1 to IB-8 + correction engine IB-5
Activate guardrails no defend + catalog only + max 2 SKU
Route ireco vs #439 product_q
Placements thumbs down T1 + widget footer T3 + feedback page T4
Structured feedback menu reason codes
Staging tests 8 scenarios: bad bot, widget, owned, correction accept, opt-out, allergy handoff, defend zero, product q route
IB-8 learning export weekly merch review calendar
Monthly ireco_bot KPI dashboard + defend audit
A/B ack copy + correction format 4 weeks
In brief
#440 = bot feedback reco tier 1, #439 agents P1 gesture
Acknowledge never defend: trust rule #1
IRECO-BOT: feedback → verify → correct grounded → log
Max 2 alternatives: no carousel post complaint
KPI ireco_correction_accept_rate: target > 55%
FAQ
Difference #439?
#439 customer service agents merch flag gesture opt-out. #440 bot ack feedback correction learning self-service.
Bot defends its suggestion?
No. Rule 1 acknowledge never defend. Always IB-6 ack first.
How many alternatives?
Max 2 SKU grounded IB-5 exclude owned OOS complained.
Allergy conflict?
bot_ireco_handoff_439 immediate P1. No alternative without metafield verify.
How does it learn?
IB-8 ireco_feedback_log weekly export merch + bot corpus review.
Going further
This week: activate thumbs down T1 on bot suggestions, configure feedback reason codes menu, test IB-5 correction exclude owned, schedule weekly IB-8 learning log review with merch.
Share this guide #440 with product and support: a bot that says "you are right, here are two suitable alternatives" is worth ten algorithm defenses, a product hallucination is worth a lost customer and a Twitter screenshot.
Product question without reco complaint?
bot_ireco_route_product_q to product questions bot, not IRECO-BOT flow.
Enzo
July 1, 2026
