Third-Party Model
A Third-Party Model is the exact payload shape one specific destination expects, derived from your Golden Model through field mapping. It is target-shaped, not source-shaped — its field names, casing, structure, and constraints match what the destination’s API demands. One Golden Model can power many Third-Party Models, each tailored to a different system.
Why use a Third-Party Model?
The Golden Model captures what your business cares about. The Third-Party Model captures what one specific system on the other end expects. Those are different concerns, and conflating them creates two recurring problems:
- Destination changes pollute your canonical schema. Klaviyo wants
$emailwith a dollar prefix. Your warehouse expectscustomer.emailnested. Your analytics tool wants flatcustomer_email. If you bake that into the Golden Model, you get a Frankenstein schema that serves nobody. - One destination’s quirks break every other delivery. When the destination team renames a field, you have to touch the Golden Model, which touches every other Third-Party Model that derives from it.
Splitting them isolates the blast radius. The Golden Model defines what the business owns. Each Third-Party Model handles one system’s quirks — renaming, restructuring, casting, dropping — without touching anything else.
How it works
Each Third-Party Model declares its own field mapping against the Golden Model. The mapping can rename fields, restructure objects, strip a literal prefix from a string, and drop fields the destination does not need. Transformation Behavior describes what a mapping does to a value and which expression forms it does not support.
graph LR G["Golden Model\n(canonical, source-shaped)"] -->|map| DM1["Third-Party Model\n(Klaviyo profile)"] G -->|map| DM2["Third-Party Model\n(Warehouse order)"] G -->|map| DM3["Third-Party Model\n(Analytics event)"] DM1 --> D1["Klaviyo"] DM2 --> D2["Warehouse API"] DM3 --> D3["Analytics"] style G fill:#a855f7,color:#fff style DM1 fill:#7c3aed,color:#fff style DM2 fill:#7c3aed,color:#fff style DM3 fill:#7c3aed,color:#fff
When a webhook arrives, it’s validated against the Golden Model first. Each Third-Party Model then runs its own mapping against the validated record to produce its destination-shaped payload. Filters can drop the delivery for any destination (e.g., “only send subscription orders to the CRM”) without affecting the others.
Example
Consider a slashbin customer’s quiz-response → email-platform pipeline. The Golden Model is a 16-field QuizResponse: respondent identifiers, answer payload, timestamps, source attribution, and a few derived classification flags.
The Third-Party Model for Klaviyo is a 22-field profile shape:
$email,$first_name,$last_name(Klaviyo’s dollar-prefixed reserved attributes)quiz_segment_v2(a computed segment string derived from three Golden Model fields)last_quiz_completed_at(Golden Modelsubmitted_at, reformatted to Klaviyo’s expected ISO-8601 with milliseconds)- A handful of
properties.*keys that namespace custom attributes the way Klaviyo expects
The Golden Model has 16 fields. The Third-Party Model has 22. The extra six come from mapping logic: derived classifications, reformatted timestamps, and Klaviyo-specific reserved attributes. If the customer adds a Postgres warehouse destination tomorrow, they author a new Third-Party Model — the Golden Model stays exactly the same.
Related concepts
- Golden Model — the canonical, source-shaped schema every Third-Party Model derives from.
- Fan-out Routing — how one Golden Model fans out to many Third-Party Models, each delivering independently.
- Getting Started — a 15-minute walkthrough covering Golden Model and Third-Party Model setup.