[ANCHOR] Customer Event-Based Engine (CEBE)

Product Requirements Document · ANCHOR v2.3 · Confluence source of truth: page 51208061912. Repo status: draft keeps the index lint-safe (anchors carry no stories); the live initiative status is ACTIVE.

Theme 1: Make the Intelligence Promise Real | Objective 1A: Strengthen Agentic AI Capabilities | H2 2026 → H1 2027 | Owner: Zhelia Alifa

This is an ANCHOR PRD. It is the master index for the CEBE initiative — it orchestrates all phases beneath it and never describes a single delivery. Per-phase execution detail (constraints, observability, dependencies, stories) lives in the child Phase / Support PRDs linked in the Phase Index.

HEADER BLOCK

Field	Value
PM	Zhelia Alifa
PRD Version	2.3
Status	ACTIVE
PRD Type	ANCHOR
Anchor	Yes — this IS the Anchor
Labels	epic:cdp | module:customer-data-platform | feature:cebe
Last Updated	2026-06-26

---

Phase Index

Phase	Goal	PRD Link	Epic Key	Status	Shipped
Phase 1 — Q3 2026: Core Infra + Customer Data + Marketing	Stand up the event-driven foundation (standardized reusable event schema, DB schema, adapter + retriever + query layer) with CDP customer-data-change events and Marketing (Broadcast + Ads) events as the first two sources; socialize to all squads by end of Q3.	(child PRDs below)	TF-3302	In Progress	—
↳ Core Infrastructure & CDP Customer Event (TECH — CDP + BI/Data lead)	Build the reusable event schema, Central DB schema, event adapter layer, retriever/metrics layer, and query approach; CDP emits customer-data-change + segment events as the first source.	prd-cebe-core-infrastructure-cdp-customer-event.md	TF-3302	Draft	—
↳ Marketing Event Source: Broadcast & Ads (SUPPORT — Broadcast squad)	Broadcast squad emits standardized Broadcast and Ads events into CEBE, enabling marketing metrics (reply rate, ads conversion) and attribution.	prd-cebe-marketing-event-source-broadcast-ads.md	TF-3302 (Owning Squad: Broadcast)	Draft	—
Phase 2 — Q4 2026: Communication, Service & Loyalty	Connect Communication (conversation, WA Call), Service (Ticket), and Loyalty events; all squads start pushing their own domain events on top of the Q3 foundation.	—	—	Not started	—
Phase 3 — Q1 2027: B2B Sales & Commerce	Connect B2B Sales (Deals), Commerce, Booking & Appointment events for full LTV coverage.	—	—	Not started	—

Note: Phase 1 uses a single shared Epic (TF-3302) covering both the Core Infrastructure (TECH) and Marketing (SUPPORT) deliverables; the Broadcast contribution is filtered via the Owning Squad field.

---

1. One-liner + Problem

One-liner: A unified, event-driven data layer that gives every Qontak squad one real-time customer profile to build segmentation, marketing automation, AI memory, and health scoring on.

Problem: Today customer data lives in silos — a ticket resolved in the support module never informs the AI Agent handling a follow-up sale, and a payment confirmed in Commerce doesn't update the customer's segmentation in real time. Each module writes its own database, producing inconsistent profiles, missed automation opportunities, and AI that cannot get full context or memory of the customer. Without a shared event-driven foundation, every new intelligence feature in H2 2026 has to re-solve cross-module data access from scratch, multiplying cost and delaying delivery across squads.

Reference data assessment: https://docs.google.com/spreadsheets/d/1KVxg-M_Ur_gW9d2DmSCysRZIgaOsoUGjlvAsiyfzklI/edit?usp=sharing

---

2. What Happens If We Don't Build This

• Every intelligence feature stalls. Agentic AI memory, segmentation, marketing automation, and customer health score all depend on CEBE — without it they cannot ship in H2 2026 and slip into 2027.
• Cross-module data work is re-paid per squad. Each squad rebuilds its own customer-data access, multiplying engineering cost and producing inconsistent profiles.
• Marketing cannot measure or target. No reusable reply-rate / ads-conversion metrics and no cross-use-case segmentation means marketing automation stays manual and un-attributable.
• AI stays context-blind. The AI Agent keeps answering without history across modules, capping CSAT and conversion gains the Intelligence theme promises.

---

3. Target Users + Persona Context

Persona	Role	Goal	Pain	Workaround
Primary — Business Admin / Marketing Owner	Business admin or marketing owner managing customer data and running campaigns in Qontak	Keep one consistent, up-to-date customer profile and target/segment customers by real behavior across channels	Customer attribute changes don't propagate; broadcast & ads engagement isn't unified, so segments and attribution are stale or impossible	Manual exports, spreadsheets, and per-channel reporting reconciled by hand
Secondary — CS / Service Agent & AI Agent	Service agent (and the AI Agent acting on their behalf) handling inbound conversations and tickets	See the customer's full cross-module history before responding, and act on risk/value signals	No memory across modules — agent/AI cannot tell an at-risk or high-value customer from a new one	Switching between modules manually to piece together context mid-conversation
Tertiary — Sales & Loyalty Owner	Sales owner (B2C/B2B deals, commerce) and loyalty program manager	Use unified deal, purchase, and loyalty signals to prioritize retention, upsell, and re-engagement	Sales, commerce, and loyalty signals are isolated from engagement and service data	Separate per-module reports that never combine into one customer view

---

4. Industry Priority

Re-prioritized 2026-06-26 to lead with the customer-data + marketing use-case, which unlocks the broadest reuse across squads first.

Priority	Use-Case / Domain	Key Use Cases
Primary	Customer Data Change & Marketing	1. Real-time propagation of customer data/attribute changes 2. Segmentation based on use-case (cross-module attributes) 3. Marketing engagement & attribution across Broadcast and Ads 4. Marketing automation / drip triggered by behavioral events 5. Reusable marketing metrics (reply rate, ads conversion)
Secondary	Consumer Service	1. Chat histories 2. Chat SLA 3. Ticket histories 4. Ticket SLA 5. Notification reminder 6. Booking completed / ticket resolution
Third	Sales (B2C & B2B) & Loyalty	1. Deal histories 2. Deal rotten time 3. Deal won 4. Commerce / purchase histories 5. Loyalty tier, points & membership powering personalized offers

---

5. Architecture — Event-Driven Customer Data Foundation

All modules publish standardized domain events to the Event Adapter Layer (contact-service), which resolves identity via qontak_customer_id and writes to the Central DB. The Retriever / Metrics Layer turns raw events into reusable metrics/summaries; downstream services read through the retriever — they never call module APIs directly.

flowchart LR
subgraph SRC[Event Sources]
direction TB
CDPMOD[CDP - Customer Data / Attributes]
CAMP[Campaign - Broadcast and Ads]
COMM[Communication - Chat, WA Call]
TICK[Ticketing - Support Tickets]
CONV[Conversion - Deals, Commerce, Booking]
LOY[Loyalty - Points, Membership]
end
ADAPTER[Event Adapter Layer - contact-service]
CEBE[(Customer Event Based Engine - Central DB)]
RET[Retriever / Metrics Layer]
subgraph OUT[Downstream Services]
direction TB
SEG[Segmentation Engine]
MA[Marketing Automation]
REP[Report and Command Center]
AI[Agentic AI Memory]
CHS[Customer Health Score]
AUTO[Automated Action]
end
SRC --> ADAPTER
ADAPTER --> CEBE
CEBE --> RET
RET --> OUT

---

6. Module Event Catalog

Each module emits standardized domain events to the Adapter. Every event carries:

(1) company_sso_id (2) company_id (3) qontak_customer_id (4) actor (5) event_timestamp (6) source_module (7) trigger_from (8) event_name (9) event_description (10) event_attributes

1. Communication Module (Chat & Call)

What it feeds CEBE: Full conversation and call history, agent interactions, channel preference, response behavior.

Event	Trigger	Key Payload Fields
conversation.started	Customer initiates an inbound chat	channel, room_id, phone, email, bsuid
call.started	Customer initiates an inbound call	channel, room_id, phone, email, bsuid
conversation.assigned	Chat assigned to human agent or bot	agent_id, assignment_type (human/bot)
call.assigned	Call assigned to human agent	agent_id, assignment_type (human/bot)
message.received	Customer sends a message	message_id, message_type, content_preview
conversation.resolved	Chat marked resolved	resolved_by, resolution_time_seconds, handled_by (human/bot)
conversation.escalated	Bot escalates to human	escalation_reason, bot_session_id
bot.contained	Bot fully resolved without human	bot_flow_id, containment_flag: true
call.completed	Call ended	duration_seconds, call_type (inbound/outbound), recording_url
csat.submitted	Customer submits CSAT	score, comment, agent_id
nps.submitted	Customer submits NPS	score, comment, agent_id

CEBE Metrics enabled: Total conversations per customer · Reply rate · Total messages sent by customer · Channel · Avg customer response time · Avg call duration · Avg CSAT · Avg NPS · AI containment rate

2. Ticketing Module (Customer Service)

What it feeds CEBE: Service history, issue resolution performance, SLA compliance.

Event	Trigger	Key Payload Fields
ticket.created	New ticket opened	ticket_id, pipeline_id, stage_id, assigned_id, creator_id, {other_fields}, customer_association
ticket.assigned	Ticket assigned to agent	ticket_id, agent_id, team_id, assigned_at
ticket.stage_changed	Ticket moves between stages	ticket_id, old_stage, new_stage, changed_by, timestamp
ticket.resolved	Ticket marked as resolved	ticket_id, resolved_at, resolution_type (first_contact/escalated), agent_id
ticket.sla_breached	SLA timer expires before resolution	ticket_id, sla_type (first_response/resolution), breach_duration_seconds

CEBE Metrics enabled per customer: Total tickets created · Resolution rate % · Avg time to resolve · Pipeline stage history · SLA breach count & breach rate · Repeat issue flag

3. Deals (B2B Sales)

What it feeds CEBE: Sales engagement depth, deal value, pipeline stage, purchase intent signals.

Event	Trigger	Key Payload Fields
deal.created	New deal opened	deal_id, pipeline_id, stage_id, deal_size, owner_id, {other_fields}, customer_association
deal.stage_changed	Deal moved to another stage	deal_id, old_stage, new_stage, days_in_stage, timestamp, changed_by, customer_association
deal.won	Deal closed won	deal_id, deal_size, close_date, {other_fields}, customer_association, product_association
deal.lost	Deal closed lost	deal_id, deal_size, lost_date, loss_reason, {other_fields}, customer_association, product_association
deal.rotten_time_breached	Rotten timer expires before conversion	deal_id, deal_size, breach_duration_seconds

CEBE Metrics enabled: Total deals created per customer · Total deal size per customer · Won rate · Lost rate · Avg time to conversion · Pipeline stage history · Rotten time breach count & rate

4. Campaign Module (Broadcast & Ads)

What it feeds CEBE: Marketing engagement history, message response behavior, and campaign attribution across both outbound Broadcast and paid Ads channels.

4a. Broadcast events

Event	Trigger	Key Payload Fields
campaign.sent	Broadcast message delivered to customer	campaign_id, template_id, channel, campaign_cost, {customer_attributes_related}, timestamp
campaign.replied	Customer replies to a broadcast	campaign_id, template_id, channel, campaign_cost, sent_at, timestamp, reply_content_preview
campaign.clicked	Customer clicks CTA in broadcast	campaign_id, template_id, channel, cta_type, destination_url, timestamp
campaign.bounced	Message not delivered	campaign_id, template_id, channel, bounce_reason
campaign.opted_in	Customer opts in to marketing	campaign_id, template_id, opt_in_channel, timestamp
campaign.opted_out	Customer opts out of marketing	campaign_id, template_id, opt_out_channel, timestamp

4b. Ads events (paid acquisition & retargeting)

Event	Trigger	Key Payload Fields
ad.impression	Ad shown to a known/resolvable customer	ad_id, campaign_id, ad_platform (meta/google/tiktok), ad_set_id, creative_id, ad_spend, timestamp
ad.clicked	Customer clicks the ad	ad_id, campaign_id, ad_platform, creative_id, destination_url, cost_per_click, timestamp
ad.lead_captured	Lead form / CTWA conversation generated from ad	ad_id, campaign_id, ad_platform, lead_form_id, room_id, timestamp
ad.converted	Ad-attributed conversion (purchase/deal/booking)	ad_id, campaign_id, ad_platform, conversion_type, conversion_value, attribution_window, timestamp
ad.spend_recorded	Daily/periodic ad spend synced from ad platform	ad_id, campaign_id, ad_platform, ad_spend, currency, period, timestamp

CEBE Metrics enabled: Total campaigns received · Delivery rate · Reply rate · Opt-in / opt-out status · Campaign-attributed revenue · Ad impressions & clicks · Ads conversion rate (% ads conversion) · Cost per lead (CPL) · Cost per acquisition (CPA) · Return on ad spend (ROAS)

5. Conversion Module (Commerce & Booking)

What it feeds CEBE: Transaction history, booking behavior, payment signals.

Event	Trigger	Key Payload Fields
booking.created	Customer creates a booking	booking_id, service_type, appointment_datetime, location_id
booking.confirmed	Booking confirmed by business	booking_id, service_type, appointment_datetime, confirmed_at
booking.cancelled	Booking cancelled	booking_id, service_type, appointment_datetime, cancelled_by, cancellation_reason
booking.completed	Service/appointment completed	booking_id, service_type, appointment_datetime, completed_at, rating
add_to_cart	Customer adds catalogue item to cart	items_id, timestamp
order.created	Order placed (Commerce)	order_id, items, total_value, payment_method
order.paid	Payment confirmed	payment_type (dp/full), paid_amount, paid_at
order.delivered	Order delivered	delivery_method, delivered_at
payment.failed	Payment attempt failed	failure_reason, attempt_count

CEBE Metrics enabled: Total bookings · Booking completed rate % · Total spend (LTV) · Avg order value · Preferred payment method · Purchase frequency · Paid rate %

6. CDP Module (Customer and Company)

What it feeds CEBE: Master identity resolution — source of truth for who the customer is. First event source connected in Q3 2026 (Phase 1).

Event	Trigger	Key Payload Fields
customer.created	New customer added (manual or automated)	qontak_customer_id, default_fields[], custom_fields[], created_at, created_by
customer.updated	Customer fields/attributes updated	qontak_customer_id, change_fields[], updated_at, updated_by
segment.entered	Customer added to a segment	qontak_customer_id, segment_id, segment_name
segment.exited	Customer removed from a segment	qontak_customer_id, segment_id, segment_name

CEBE Metrics enabled: Total new customers · New customer growth rate · Customer acquisition by source · Segment per customer

7. Loyalty Module

What it feeds CEBE: Retention signals, member behavior, tier progression.

Event	Trigger	Key Payload Fields
membership.enrolled	Customer joins loyalty program	qontak_customer_id, member_id, program_id, enrolled_at, initial_tier, available_point
membership.unenrolled	Customer unenrolls from program	qontak_customer_id, member_id, program_id, unenrolled_at, last_tier, available_point
points.earned	Points credited	qontak_customer_id, member_id, points_amount, transaction_ref, earned_at, available_point, expired_point_at
points.burned	Points burned for reward / deduction	qontak_customer_id, member_id, points_used, redeemed_at, available_point, expired_point_at
rewards.redeemed	Reward redeemed	qontak_customer_id, member_id, reward_used_id, reward_name, available_point, expired_point_at
tier.upgraded	Customer reaches higher tier	qontak_customer_id, member_id, old_tier, new_tier, available_point, expired_point_at
tier.downgraded	Customer drops tier	qontak_customer_id, member_id, old_tier, new_tier, available_point, expired_point_at

CEBE Metrics enabled: Total loyalty members · Total available points per member · Enrollment rate · Enrollment by program · New member growth rate · Tier distribution

---

7. CEBE as Baseline for New Features

Downstream Feature	Dependency on CEBE
Customer Segmentation	Rule engine filters customers by cross-module attributes (e.g. "2+ tickets AND no purchase in 90 days").
Marketing Automation / Drip	Triggers on behavioral events (e.g. auto re-engage when booking.cancelled or ad.lead_captured fires).
Agentic AI Memory	AI reads full customer history before responding. Without CEBE, AI has no cross-module memory.
Customer Health Score	Score computed from ticket volume, CSAT trend, deal stage, loyalty tier, last engagement date.
Report & Command Center	Manager dashboard shows per-customer timeline, cross-module metrics, AI-generated insights.
Automated Action	Rule-based triggers fire when CEBE detects risk signals (e.g. SLA breach → auto-escalate → notify supervisor).

---

8. Proposed Timeline & Priority

Q	Goals	Detail Scope	DoD	Child PRD	Dependency
Q3 2026	Core Infrastructure, Customer Data, Segment & Marketing — P0 Blocker	• RFC & assessment, both sides (Engineering & Data). • Standardize a reusable event schema for all squads (push like Mixpanel, keyed on qontak_customer_id via ContactResolver). • Provide the database schema (event-level + customer-level aggregation). • CDP starts providing customer events — see 6. CDP Module. • Marketing use-case — see 4. Campaign Module; connect Broadcast and Ads (not Broadcast only). • Build the event adapter layer (contact-service) — identity resolution + idempotency + PII masking. • Build the retriever layer (metrics/summaries, e.g. % reply rate, % ads conversion). • Provide the query approach for downstream features (segmentation, marketing automation). • Socialize to all squads at end of Q3 so Q4 every squad can push their domain events.	• RFC approved by both Engineering and Data. • Standardized event schema published and adopted by CDP + Broadcast. • Central DB schema + event adapter + retriever + query layer live. • 100% of CDP customer events and Marketing (Broadcast + Ads) events propagated to CEBE in <5s. • Reusable metrics queryable for segmentation/marketing automation. • All squads socialized and able to push their domain events in Q4.	Core Infrastructure & CDP Customer Event — TECH · Marketing Event Source: Broadcast & Ads — SUPPORT	Data · CDP · Chat 2 (Broadcast)
Q4 2026	Communication, Consumer Service & Loyalty	• Communication (conversation, WA Call) — see 1. Communication Module. • Consumer Service (Ticket) — see 2. Ticketing Module. • Loyalty — see 7. Loyalty Module.	• 100% of Communication, Ticket, and Loyalty events propagated to CEBE in <5s.	—	Data · CDP · Chat 1 (Comms) · Voice (WA Call) · CRM (Ticket) · Loyalty
Q1 2027	B2B Sales (Deals), Commerce, Booking & Appointment for complete LTV coverage	• B2B Sales (Deals) — see 3. Deals (B2B Sales). • Commerce, Booking & Appointment — see 5. Conversion Module.	• 100% of Deal, Commerce, Booking & Appointment events propagated to CEBE in <5s. • Per-customer LTV and Customer Health Score live for all active accounts.	—	Data · CDP · CVT · CRM (Deal) · Innovation (Booking, etc)

Detail Task per Phase

Phase 1 — Q3 2026

Task	Owner	Timeline
RFC & feasibility assessment (Engineering + BI/Data)	PM, Eng (CDP), BI/Data	Jul 2026
Design standardized reusable event schema + qontak_customer_id idempotency	BI/Data, PM, Eng (CDP)	Jul 2026
Provide Central DB schema (event + aggregation layer)	Eng (CDP), BI/Data	Jul 2026
Build Event Adapter Layer (contact-service)	Eng (CDP)	Jul–Aug 2026
Connect CDP customer-data-change + segment events	Eng (CDP)	Aug 2026
Connect Marketing events — Broadcast + Ads	Eng (Broadcast)	Aug 2026
Build retriever / metrics layer (reply rate, ads conversion, etc.)	Eng (CDP), BI/Data	Aug 2026
Provide query approach for downstream features (segmentation, MA)	Eng (CDP), BI/Data	Aug–Sep 2026
Internal monitoring & alerting (<5s SLA)	Eng (Infra)	Aug 2026
Socialize foundation + event-push guide to all squads	PM, CDP	End of Sep 2026

Phase 2 — Q4 2026

Task	Owner	Timeline
Connect Communication events (conversation, WA Call)	Eng (Chat & WA Call)	Q4 2026
Connect Ticketing events	Eng (Ticket)	Q4 2026
Connect Loyalty events	Eng (Loyalty)	Q4 2026
Agentic AI reads CEBE memory	Eng (Chatbot)	Q4 2026

Phase 3 — Q1 2027

Task	Owner	Timeline
Connect Deal events	Eng (Deal)	Q1 2027
Connect Commerce events	Eng (Commerce)	Q1 2027
Connect Booking & Appointment events	Eng (Booking)	Q1 2027
Build Customer Health Score model	Eng (AI/Data)	Q1 2027

---

9. Technical Principles

• Event-Driven, Not Polling: All modules publish events asynchronously. CEBE never calls module APIs directly.
• Standardized, reusable event schema: One event contract for all squads — push customer events like Mixpanel, keyed on qontak_customer_id via ContactResolver.
• contact-service is the authoritative write service: Identity resolution and deduplication happen here before data lands in Central DB.
• Idempotency via qontak_customer_id: No duplicate events regardless of retry or replay.
• < 5s propagation SLA: Monitored via internal alerting; any breach triggers on-call notification.
• Backfill on first integration: When a module first connects, a backfill job loads the last 12 months of historical events.
• PII-aware: Sensitive fields are masked at the adapter layer based on role-based access policy.

---

10. Success Metrics (Initiative-level)

Category	Metric	Definition	Baseline	Target
Adoption & Usage	⭐ Module event coverage on CEBE	% of target modules emitting standardized events into CEBE	0% (no shared layer today)	CDP + Marketing (Broadcast+Ads) by end of Q3 2026; Comms/Ticket/Loyalty by end of Q4 2026; full coverage by end of Q1 2027
Quality & Accuracy	Event propagation latency	Time from source event to availability in Central DB (p99)	N/A — new	< 5 seconds (p99) from end of Phase 1
Quality & Accuracy	Zero event loss	Lost events across module streams	N/A	0 lost events (idempotency via qontak_customer_id) — ongoing
Efficiency & Impact	Reusable metrics & segments available	Reply rate, ads conversion, use-case segments queryable from CEBE	N/A	Available for segmentation + marketing automation by end of Q3 2026
Efficiency & Impact	AI Agent uses CEBE memory	% of conversations where AI reads CEBE history before responding	0%	≥ 50% of conversations by end of Q4 2026
Efficiency & Impact	Customer Health Score live	Active accounts with a computed health score	0	All active accounts scored by end of Q1 2027

---

11. Key Decisions + Alternatives Rejected

11a — Decisions Made

Date	Decision	Rationale
2026-06-26	Lead Q3 with the Customer Data + Marketing use-case (was: Consumer Service)	Broadest, earliest reuse across squads; unlocks segmentation + marketing automation first
2026-06-26	Marketing scope includes Ads, not Broadcast only	Ads attribution (CPL, CPA, ROAS, % ads conversion) is core to the marketing use-case
2026-06-26	Q3 delivers a standardized, reusable event schema any squad can push to	Squads self-serve event push (like Mixpanel) keyed on qontak_customer_id; avoids per-squad re-implementation
2026-06-26	Phasing: Q3 Core+CDP+Marketing → Q4 Comms/Service/Loyalty → Q1'27 B2B Sales/Commerce	Sequences by reuse value and squad readiness; socialize at end of Q3 so all squads onboard in Q4
2026-06-26	Q3 owned jointly by CDP + BI/Data	RFC & assessment required on both engineering and data sides; retriever/query layer is data-heavy
2026-06-26	Phase 1 uses a single shared Epic (TF-3302) for both TECH + SUPPORT	One delivery slice; Broadcast contribution filtered via Owning Squad rather than a separate Epic

11b — Alternatives Rejected

Alternative	Why Rejected	Date
Each module keeps its own DB; build point-to-point integrations	Re-pays cross-module data work per squad; inconsistent profiles; doesn't scale to AI/segmentation	2026-06-26
Polling module APIs on demand instead of event-driven	Cannot meet <5s freshness; tight coupling; high load on module APIs	2026-06-26
Lead Q3 with Consumer Service (previous priority)	Narrower early reuse than customer-data + marketing; delays segmentation/MA foundation	2026-06-26
Broadcast-only marketing scope	Misses paid-acquisition attribution; cannot compute ads conversion / ROAS	2026-06-26

---

12. Open Questions

#	Type	Question	Owner	Deadline
1	Open Question	Event-level (append-only log) vs aggregated tables vs hybrid for the Central DB?	BI/Data + Eng (CDP)	2026-07-15
2	Open Question	Do we need streaming infra (Kafka / PubSub) to guarantee <5s and zero loss, or is the current pipeline sufficient?	Eng (CDP) + Infra	2026-07-15
3	Open Question	Final query approach for downstream consumption (pre-aggregated vs on-the-fly)?	BI/Data + Eng (CDP)	2026-08-15
4	Assumption	All squads can adopt the standardized event schema in Q4 after end-of-Q3 socialization	PM (CDP)	2026-09-30
5	Risk	Module teams slow to emit events → downstream delayed. Mitigation: phase-gate by module; CDP + Marketing first in Q3; socialize + push guide before Q4	PM (CDP)	2026-09-30
6	Risk	High write volume causes >5s breach. Mitigation: pre-size DB; async queue buffer between adapter and DB	Eng (Infra)	2026-08-31
7	Risk	PII leakage across team boundaries. Mitigation: PII masking at adapter + team-level permission scope	Eng (CDP)	2026-08-31
8	Risk	Duplicate identities / late-arriving events. Mitigation: idempotency + dedup on qontak_customer_id at adapter	Eng (CDP)	2026-08-31

---

13. Appendix — Data Assessment Request

Retained from the original CEBE assessment brief — input to the Q3 RFC & assessment workstream.

Request Title	Data Assessment for CEBE (Customer Event-Based Engine)
Request Objective	Build CEBE as a unified, customer event-driven data layer supporting Customer Insight (aggregated summary per customer) and Conversation AI (profiling, sentiment, behavior). Assess real-time event ingestion + cross-module query feasibility; evaluate a centralized event-driven data model; identify need for new datamarts, streaming pipelines, or storage optimization.
Data Usability Plan	A. Customer Insight / Customer 360 — unified profile combining Identity (CDP), Engagement (Chat, Campaign/Ads), Service (Ticket), Sales (Deal, Commerce), Loyalty. B. Conversation AI — real-time + historical behavioral signals for AI decisioning and context-aware responses. C. Downstream — marketing automation triggers, segmentation engine, cross-module reporting.
Key Requirements	A. Data model & storage — event-level vs aggregated vs hybrid; <5s ingestion + analytical queries. B. Ingestion — high-volume streams; <5s, idempotency, 12-month backfill, zero loss. C. Query — aggregated customer query (<1–2s) + granular event/AI query. D. Identity — all events linked via qontak_customer_id; handle duplicates, late events. E. Metrics — engagement/service/sales/loyalty/commerce; precomputed vs on-demand. F. AI-ready layer — structured + semi-structured; feature store need. G. Governance — retention, partitioning, PII access control, storage growth & cost.

---

PRD CHANGELOG

Version	Date	By	Section	Type	Summary
2.0	2026-06-26	Claude	All	MODIFIED	Converted to ANCHOR PRD; updated Industry Priority; re-cut timeline (Q3/Q4/Q1); added Ads to Campaign Module; added retriever/query layer.
2.1	2026-06-26	Claude	Phase Index, S8	MODIFIED	Linked the two Phase 1 child PRDs; fixed ToC + anchors.
2.2	2026-06-26	Claude	S1	MODIFIED	Tightened one-liner to ≤25 words.
2.3	2026-06-26	Zhelia Alifa / Claude	S5, S6, S8	MODIFIED	Linked Epic TF-3302; reworked S8 into table + per-phase task tables; expanded event envelope to event_name/description/attributes; architecture flow made linear (Central DB → Retriever → Downstream). Committed to documents repo as .md.