Skip to main content
Recent, customer-facing changes to SourceMedium and to this documentation.
Dates on this page are when documentation was published, not when a capability shipped. A feature can go live before its guide lands here. For the live state of your own data sources, open your workspace and check connection statuses.
​
Agent skill for building your own data pipelines

​
Build your own pipelines next to SourceMedium data

A third agent skill, SM Pipeline Builder, is available for technical teams who want to load a source SourceMedium does not connect to into their own BigQuery datasets and join it to SourceMedium data.The skill writes a validated pipeline spec before any code: grain, primary key, cursor and overlap window, first-build window, delete behavior, and the checks that run on a schedule. It measures the source with a small live probe rather than trusting vendor docs, loads incrementally so any window can be safely re-run, and reads SourceMedium datasets without ever writing to them.Pipelines built this way run in your infrastructure. SourceMedium-managed connectors are unchanged, and a managed connector that stops loading is still a support request.
​
Faster multi-table AI assistant discovery

​
Faster multi-table discovery

Connected assistants now have ten read-only tools. The new describe_tables tool compares two to four warehouse tables in one request, helping assistants plan joins from real schemas without repeating one-table lookups. See Connect an AI assistant and the MCP tool reference.
​
AI assistant setup and tools, media mix tables

​
AI assistants

Connect an AI assistant now carries step-by-step setup for each supported tool: Claude, ChatGPT, Claude Code, Codex, Cursor, Grok Build, and Hermes. The steps match the AI assistants page inside your SourceMedium workspace.At publication, a connected assistant got nine tools. New since the original seven: get_account_health answers “are my connections working?” with per-connection status and data freshness, and search_docs searches this documentation so the assistant can explain SourceMedium concepts instead of guessing. The renamed MCP tool reference covered those nine.Shopify and Meta no longer need a separate review. Each answers as soon as the store or ad account is connected to SourceMedium; until then the tool returns SOURCE_UNAVAILABLE and warehouse questions keep working.

​
Media mix tables

Pro warehouses now document two media mix tables in sm_transformed_v2: rpt_media_mix_channel_daily (start here) and obt_media_mix_performance (campaign and ad detail, plus which sales venue the revenue landed on).These tables are Pro only. Foundation still receives 16 tables; Pro receives all 33. Start from the daily report when comparing modelled, last-click, and platform-reported ROAS. See the Fospha integration guide and Foundation and Pro.
​
Shopify reporting coverage

​
Shopify reporting coverage

SourceMedium expanded its Shopify reporting model for eligible stores. Product and collection metadata now use the current Shopify catalog walk, fulfillment status can use Shopify fulfillment headers and line events, and customer-journey moments can fill last-touch attribution gaps when stronger evidence is unavailable.Native Shopify Return records can now populate fct_returns when a store does not use Loop Returns or ReturnGO. If either returns provider has landed data, native Shopify returns are excluded to prevent duplicate returns. Fields Shopify does not provide, such as complete return money and return-level timestamps, remain NULL.These changes roll out through the normal scheduled transformation and warehouse refreshes. See the Shopify integration guide, fct_returns, and the Attribution Source Hierarchy.
​
AI assistants, warehouse plans, and Connect

​
AI assistants

You can connect any MCP-capable assistant to SourceMedium. Copy the connection URL from the AI assistants page in your workspace, or use https://mcp.sourcemedium.com/mcp. ChatGPT, Claude, and Grok are typical; the same URL works in other MCP clients. Warehouse access is on for every active organization. Shopify and Meta need an extra review.MCP tool reference lists the tools an assistant gets, the call order that produces correct numbers, and the column name assistants most often guess wrong.The Slack AI Analyst now documents Slack Connect store intersection, streaming replies, and native Good/Bad feedback. Understanding results now states the row caps: 5,000 for an aggregated analysis, 100 for a raw data pull.

​
Warehouse

Foundation is the shared warehouse plan. Pro is the dedicated project. Foundation and Pro now names exactly which 17 tables are Pro only, and each of those table pages carries a banner. The catalog adds dim_subscribers, fct_subscription_charges, and obt_subscription_charges, plus obt_affiliate_conversions on the index. Funnel events list Elevar, RudderStack, Snowplow, and GA4.

​
Connect

ShopMy, Superfiliate, Judge.me, Skio, and Stay AI now connect with API credentials; each guide walks through creating them. X Ads is launched. ChatGPT Ads now uses the monitored connection-status flow after you save a valid key. The GA4 guide now documents what the self-serve OAuth connection syncs: sessions, events, transactions, and item-level ecommerce from the GA4 Data API.
​
Documentation

​
Docs reorganized by job rather than product structure

The tabs now map to what you are trying to do:
  • Workspace for setup, access, data sources, and the daily Slack report
  • Connect for integrations, the configuration sheet, tracking setup, and health checks
  • How your data works for how we clean, enrich, and attribute your data, wherever you read it
  • Dashboards for Looker Studio, modules, and templates
  • Warehouse for BigQuery, SQL, table docs, and metric definitions
  • MTA, AI Analyst, and Help unchanged in purpose
Page URLs did not change, so existing bookmarks and shared links keep working.

​
Also added

  • Reporting currency behavior across order, ads, and subscription surfaces
  • Customer-facing schema pages for the v2 subscription tables
​
Integrations and workspace guides

​
New integrations

​
Workspace guides refreshed

Every workspace product tour was recaptured from the current UI, including dark mode. The post-payment flow now hands you straight to connecting your first data source rather than to Slack setup.

​
Warehouse

Added the ticket_api_path column to the obt_customer_support_tickets schema docs.

​
Fix

The funnel mapping table listed page as the RudderStack source event. The pipeline matches on pages, which is what you will find in your dataset.
​
Integrations

​
ChatGPT Ads

Added the ChatGPT Ads integration guide. Connect with an Ads API key from OpenAI Ads Manager.
​
Warehouse and data accuracy

​
New table documentation

  • sm_views dataset docs, including rpt_customers_first_and_last_order_summary_v1 and rpt_order_returns_v1
  • rpt_affiliate_performance_daily, with Superfiliate marked live

​
Definitions clarified

  • Amazon data settling curve and freshness windows
  • Amazon Subscribe & Save order classification
  • Fraud and declined cancellations in order validity rules
  • Exclude-$0 filtering keys on net revenue, not total revenue
​
AI Analyst

​
AI Analyst workflows and agent skills

Added workflow pages for raw data pulls and for summaries and opinions, plus documentation for the SourceMedium agent skills. See AI Analyst workflows and Agent skills.