The Closed Loop — From Recommendation to Outcome

What It Is

The Closed Loop is Korvex's fully automated pipeline from recommendation to measured outcome. Unlike traditional SEO workflows where recommendations sit in spreadsheets, Korvex's closed loop connects every stage: SCIS generates a recommendation → you approve it → the CMS bridge deploys the change to your website → the outcome tracker measures the impact → that outcome calibrates future predictions. It's a self-improving system.

Why It Matters for Your SEO

The #1 failure mode in SEO is the "recommendation gap" — insights that never get implemented, and implementations that never get measured. The closed loop solves both:

No implementation gap: Approved recommendations are automatically pushed to your CMS (WordPress, Shopify, or Webflow)
No measurement gap: Changes are tracked from deployment through to ranking/traffic outcome
Self-improving accuracy: Every measured outcome makes future predictions more accurate

How Korvex Measures It

The Lifecycle

SCIS Recommendation → User Approval → CMS Deployment → Outcome Measurement → Prediction Calibration
      (pending)        (in_progress)     (deployed)        (completed)           (feedback loop)

Workflow Statuses

Status	Meaning	How It's Set
Pending	Backlog — awaiting prioritisation	Default on creation
In Progress	Being worked on — baseline metrics captured	User drags card on Strategy Board
Deployed	Live on website, being monitored	Auto-promoted by CMS executor
Completed	Outcome measured, proven effective	Auto-promoted by outcome tracker
Dismissed	Won't implement	User drags card
Rolled Back	Was deployed, then reverted	CMS rollback action
Failed	Implementation attempted but errored	CMS executor error

Auto-Promotion Rules

In Progress → Deployed: Automatic when CMS executor successfully pushes changes. Sets deployed_at.
Deployed → Completed: Automatic when daily outcome tracker (12:00 UTC) measures success. Requires time_to_impact_days to have elapsed since implementation.
Dismissed → Unblocks dependents: Dismissing a recommendation removes it from dependent recommendations' blocking lists.

How to Improve Your Score

Connect your CMS — link WordPress, Shopify, or Webflow in Client Settings for automated deployment
Approve recommendations promptly — the sooner changes deploy, the sooner outcomes are measured
Don't skip the board — drag cards through the kanban stages to ensure proper lifecycle tracking
Review the timeline — the Strategy Timeline shows causal relationships between changes and outcomes
Be patient with outcomes — impact measurement requires time_to_impact_days to elapse (typically 14-30 days)

<details> <summary>Technical Deep Dive</summary>

CMS Executor Configuration

Task: execute_recommendation on queue="default"
Time limits: 9 minutes soft, 10 minutes hard
Retries: 3 attempts with exponential backoff (5s → 10s → 20s, max 300s)
Error classification: RetryableError (network, rate limit) → retry; PermanentError (invalid credentials) → mark failed

Execution Strategies

Type	Strategy	Complexity
schema_addition	HTML parsing	Complex — injects JSON-LD
heading_restructure	HTML parsing	Complex — modifies DOM
content_refresh	HTML parsing	Complex — replaces body content
internal_linking	HTML parsing	Complex — inserts anchor tags
All others	Field mapping	Simple — updates CMS fields directly

Execution Workflow (7 Steps)

Load recommendation + CMS mapping (multi-tenant filter)
Guard: skip beacon-channel recommendations (delivered client-side)
Set status to implementing, record implementation_started_at
Determine strategy: ACF generation, HTML parsing, or field mapping
Execute CMS API call (WordPress REST, Shopify Admin, or Webflow API)
Create version snapshot (audit trail)
Update status to implemented, auto-promote workflow_status to deployed

Outcome Tracker

The daily outcome tracker (12:00 UTC):

Compares current metrics to baseline captured at implementation
Evaluates: ranking change, traffic change, CTR change, Koray score change
Outcome classification: success, partial_success, no_change, negative
Auto-promotes workflow_status to completed when outcome is success/partial_success

Key Tables

page_recommendations — the primary recommendation store
cms_recommendations — CMS-specific execution details
cms_page_mappings — maps recommendation page IDs to CMS page IDs

Key Files

Transitions: services/api/services/workflow_orchestrator.py
CMS execution: services/cms/recommendation_executor.py
Outcome measurement: services/analyzers/recommendation_outcome_tracker.py
Frontend: frontend/src/utils/strategyConstants.ts

SCIS — generates the recommendations that enter the loop
Confidence and Impact — outcomes calibrate future predictions
Content Briefs — the specification for what gets deployed

</details>