Finance Attribution Model

Vision uses a 3-tier identity hierarchy to attribute subscription revenue and customer lifetime value (LTV) to earlier marketing and user journeys.

The 3-Tier Identity Hierarchy

+--------------------------------------------------------------------------+
| Tier 1: Canonical userId (Identified via X-Vision-User-Id / auth.id)      |
|         Highest confidence. Directly links Stripe customer & LLM tokens. |
+--------------------------------------------------------------------------+
                                     │ (fallback if unauthenticated at checkout)
                                     ▼
+--------------------------------------------------------------------------+
| Tier 2: visitorId (Persistent UUID in localStorage / client tracker)     |
|         Links pre-login marketing traffic to payment checkout.          |
+--------------------------------------------------------------------------+
                                     │ (fallback if cross-device / cookie cleared)
                                     ▼
+--------------------------------------------------------------------------+
| Tier 3: sessionId / deviceId (Deterministic TLS & Header Fingerprint)    |
|         Probabilistic fallback for guest checkouts.                      |
+--------------------------------------------------------------------------+

1. Tier 1: Identified userId

  • Trigger: Triggered via vision.identify({ userId }) or the X-Vision-User-Id HTTP header.
  • Accuracy: 100% deterministic.
  • Use case: Reconciles recurring invoice renewals, team expansion seats, and ongoing model inference usage.

2. Tier 2: Persistent visitorId

  • Trigger: Generated by the lightweight tracker script and saved in client storage.
  • Accuracy: Deterministic within browser.
  • Use case: When an anonymous visitor reads documentation, signs up for a trial, and completes a payment checkout.

3. Tier 3: Deterministic deviceId

  • Trigger: Synthesized from request IP, User-Agent, and daily cryptographic salt without cookies.
  • Accuracy: Probabilistic.
  • Use case: Handles cross-tab or cross-session guest checkouts where cookies or local storage are blocked.

Multi-Touch Revenue Attribution

Vision attributes conversion revenue using two models:

  • First-Touch Attribution: Attributes 100% of the initial subscription value to the original referrer or UTM campaign that first introduced the visitor.
  • Last-Touch Model Attribution: Attributes renewal and expansion MRR to the specific AI feature or prompt workflow that triggered the upgrade prompt.