How Does SPAN work?

SPAN is a Snowflake-native identity and semantic modeling platform.

It resolves fragmented records into governed identity profiles and transforms them into standardized, business-ready dimensions — entirely inside your data warehouse.

SPAN consists of two core modules:

  • ID Graph — Identity resolution

  • Compiler — Semantic layer for metrics and segments

Identity becomes transparent, versioned, and configuration-driven.

Identity Graph

The ID Graph resolves records from multiple systems (e.g., Shopify, Ticketmaster, CRM) into unified profiles.

Each profile is assigned a stable profile_id.

Deterministic Blocking Rules

Identity resolution is defined explicitly in configuration.

Examples:

  • Email + Last Name

  • Phone + Zip Code

  • First Name + Last Name + Birthdate

Each rule generates clusters of related records. Overlapping clusters merge into a single identity.

There is no hidden matching logic.

Overrides

SPAN includes an explicit Override Table to manage edge cases.

Users can:

  • Merge records

  • Split records

  • Assign orphaned records

Overrides are reapplied automatically every time the graph runs.

Snapshots & Lineage

The ID Graph stores historical states.

You can:

  • Compare graph versions

  • Inspect cluster changes

  • Audit identity evolution over time

Identity is versioned and queryable.


The Compiler

The Compiler transforms resolved profiles into structured business attributes.

All behavior is defined in configuration (e.g., c360_config.yml).

There are no hardcoded SQL scripts.

What the Compiler Produces

Four profile-linked dimensions:

  • Demographics — Name, Email, Phone, Address

  • Metrics — LTV, Orders, Engagement

  • Segments — Cohorts and flags

  • Consent — Unified marketing permissions

All outputs are keyed by profile_id.

Precedence Logic

When attributes conflict across systems, SPAN applies configurable rules.

Examples:

  • Prefer Shopify shipping address over CRM

  • Take the greatest value for Last Active Date

  • Compute Total LTV from component metrics

Attribute logic remains explicit and auditable.


DevOps for Identity

SPAN introduces structured change management to identity logic.

  • Rules are defined in configuration

  • Changes are tested before promotion

  • Diff reports show impact before deployment

This prevents unintended merges and silent logic drift.


Where SPAN Fits

SPAN operates in the Silver Layer of the data stack.

It:

  • Resolves identity

  • Standardizes attributes

  • Produces governed profile tables

Downstream tools consume SPAN outputs for analytics, marketing, and AI workflows.


In Summary

SPAN makes identity:

  • Deterministic

  • Transparent

  • Versioned

  • Config-driven

  • Native to Snowflake

Identity becomes infrastructure and not a black box.

Last updated