v3.9 · Iteration Jul 6, 2026 → Jul 19, 2026
DeepChatBI Product Evolution — Version 3.9
Ad-decision focus restored: TikTok & Meta ontology into Profit World Model actions, long-lived Facebook auth, DataLab isolation, pixel/ChatBI hardening, and LinkedIn-led GTM outbound—while Klaviyo recall moves to maintenance.
Version 3.9 recenters the product on a single decision spine: Google, Meta, and TikTok ad actions that operators can ship—while lifecycle (Klaviyo) stays valuable but no longer splits engineering focus.
Strategic Milestone: One Spine, Not Two Fronts
After shipping Klaviyo Profile Recall and Flow-adjacent agents earlier in the cycle, we paused further retention feature expansion. Maintenance continues; new R&D capacity returns to advertising decision agents.
Operating principle this sprint:
- Customer demand → white-glove delivery → reusable warehouse & agent assets → repeated validation → PMF → standardized delivery
- Agent-assisted workflows across CRM: ontology modeling, AI ETL health, PRD design, and LinkedIn outreach automation
The benefit: The team ships one coherent Ad Decision OS instead of splitting capacity across two unfinished PMFs.
TikTok + Meta Ontology → Actions in Profit World Model
Paid-media ontology for TikTok Ads and Facebook/Meta Ads moved from design to executable product surface.
Warehouse & ontology
- Dimension tables and source-table foundations for TikTok advertising
- Shared decision vocabulary so Meta/TikTok entities map cleanly into cards and action reports
- Design docs → ontology assets → configured cards → Action reports surfaced in Profit World Model
Agent & backend actions
- Google Ads action path continued hardening
- Backend Meta / TikTok action operations added to the agent analysis stack
- TikTok skills + SQL generation for analysis workflows
- Algorithm removal / decision results written to MySQL where required for execution loops
The shift: TikTok and Meta are no longer “connectors only”—they enter the same traffic-light action loop as Google.
Facebook Auth Hardening (Long-Lived Tokens)
User 60-day re-auth was a reliability cliff for Ads → BigQuery → ChatBI.
What we shipped / decided:
- Early expiry notices (≈7 days) with email + in-app messaging so operators re-authorize before pipelines go dark
- Architecture guidance: prefer Business Manager System User tokens for the Ads → warehouse → ChatBI path instead of short-lived user tokens
The benefit: Attribution and budget recommendations stay online when individual logins would have expired silently.
Pixel, ChatBI & Tracking Fidelity
Ad decisions still depend on event truth. This cycle closed several production gaps.
Pixel & journey completeness
- Core-path pixel coverage expanded; FE/BE tracking fields enriched (
visit_type,visit_name) - Journey gaps vs lifecycle platforms reviewed; deployment path for the next pixel package prepared
- Alignment work started so GA4 and first-party pixel can cross-check funnels (checkout friction, drop-off)
ChatBI quality
- Chat history filters intermediate agent artifacts so operators see answers—not raw tool logs
- Guardrails to keep customer-facing chat surfaces in English (no leaked intermediate-language noise)
- Recall / RACell configuration surfaces for retention-adjacent diagnostics where still needed for design partners
DataLab Isolation, Notifications & Commerce Ops
DataLab FE split
- DataLab frontend separated and deployed with data-layer isolation
- Production and test hosts for safer admin and pipeline operations
In-app messaging & billing
- In-app notification FE/BE shipped for ops alerts (e.g., Meta token expiry)
- Payment success and renewal emails; Shopify trial / install deep-link parameters hardened (
email=on App Store install flows) - RetentionOS materials submitted for review
Cost & pipeline ops
- BigQuery frequency control and CRM cost visibility to keep warehouse spend predictable
- AI Pipeline: BigQuery-backed logs/results, failure retries, auto start/stop, workflow split by flow type
Multi-Channel Connectors & CRM Outbound
Connector progress:
- Snapchat integration completed earlier in the window; LinkedIn / Twitter permission and attribution work continued
- Amazon Ads (NA) remained in pre-release / translation path toward listing
- TikTok attribution decision design docs with mapping to live campaign tables
CRM & GTM infrastructure:
- Linked API workspace for target ICPs, conversation sync, and post history (ops can review results; search remains device-bound)
- Merchant directory / QDC refactors; FetchTask removal for cleaner CRM jobs
- Cold-email / GTM mail pipelines iterated (Klaviyo naming, send visibility, daily automation checks)
- Upwork + DTC pipeline active for white-glove analytics engagements that feed back into product assets
Official Site & Social Presence
- Marketing site pricing footer and conversion UX polish
- LinkedIn company page live with first content batch; X/Twitter reply + list strategy after Communities sunset
- Historical ICP engagement on social as a seeding channel—not spray-and-pray
Team & Focus
- Warehouse + FE capacity onboarded to own ad-journey code, TikTok/Meta surfaces, and ChatBI reliability
- Community / social operations launched with LinkedIn + X as primary channels
- Hiring signal: commercial/GTM and data-warehouse capacity remain open roles so founders stay on traction and investor updates
What's Coming in Version 3.10
- TikTok + Meta Action reports fully live in Profit World Model cards for daily operators
- System User token path as default for Meta Ads ingestion
- Amazon Ads (NA) production attribution in the decision loop
- Cold email + LinkedIn outreach with measurable weekly send/reply attribution
- Pixel ↔ GA4 reconciliation views for checkout drop-off questions
Core Vision
DeepChatBI is the ad decision OS for Shopify.
One spine: store truth + paid media (Google, Meta, TikTok) + enough lifecycle context to rank the budget move you ship today.