SocialHub.AI
← Data Architecture
Sources & collection

Ingestion is how every retail signal gets in, without a single fragile pipe.

The first layer is responsible only for collection. It brings retail data in through four complementary paths: enterprise-grade product adapters for CDP, CRM, ERP, OMS and commerce systems; zero-copy federation that queries data in place in Snowflake, Databricks, Salesforce Data Cloud or Delta Sharing; behavior and event collectors that capture web and app activity through an SDK; and an open API/SDK data interface for anything the packaged paths do not cover.

Layer
Ingestion

Data enters four ways: curated product adapters, zero-copy federation, behavior/event collectors and an open data-interface API/SDK.

Why it matters

The value is operational, not just technical.

Because ingestion is a distinct layer, the AI never depends on how a given source happens to be connected. New systems can be added through an adapter, federated in place, or sent through the open interface, and everything downstream sees the same kind of incoming data.

Offers four collection paths so teams pick the right one per source instead of forcing one method.

Keeps enterprise systems on maintained adapters rather than bespoke one-off integrations.

Supports zero-copy federation so large warehouse datasets can be used without duplicating them.

Provides an open API/SDK interface for custom sources the packaged connectors do not cover.

Architecture contract

What this layer is responsible for

1

Curated product adapters connect CDP, CRM, ERP, OMS and commerce platforms with mapped schemas.

2

Zero-copy federation reads data in place from Snowflake, Databricks, Salesforce Data Cloud and Delta Sharing.

3

Behavior collectors capture web and app events through an SDK at the point of interaction.

4

An open data-interface API/SDK accepts custom sources that the packaged paths do not handle.

Retail signal coverage

What it makes usable

member, order and inventory records from CRM, ERP, OMS and commerce systemswarehouse tables federated in place from Snowflake, Databricks or Data Cloudweb and app behavior: page views, product views, search, add-to-cart and clickscustom sources pushed through the open API/SDK interface
Agency AI impact

This is how the layer improves AI decisions.

Without a dedicated ingestion layer, every source becomes a bespoke pipe. That leaves teams maintaining brittle one-off integrations, copying data they did not need to move, and unable to add a source without touching everything downstream.

A commerce platform can be connected through its adapter rather than a hand-built integration.

A large warehouse dataset can be queried in place through federation instead of being copied first.

A storefront's browse and cart events can be captured with the SDK and arrive alongside order data.

Continue the architecture

Real-time capability comes from the complete chain.