Orchard <> Magicspoon

Date: February 17, 2026 Source: Granola Meeting ID: e928df2d-3447-458e-9b3e-002c0c9b6371 URL: https://notes.granola.ai/t/e928df2d-3447-458e-9b3e-002c0c9b6371

Participants:

  • Uttam Kumaran (Brainforge)
  • Ryan Brennan (Skio/Orchard Analytics)
  • Roland Cassirer (TireAgent/Orchard)
  • Anne (Magicspoon)
  • Mary (Magicspoon)
  • Nelle Lightbourn (Magicspoon)
  • Amelia (Magicspoon)
  • Alexa Minotti (Magic Spoon)
  • Justin Tabarini (Magicspoon)
  • Alejandro (Magicspoon)
  • Michael Thorson (Magicspoon)
  • Demilade Agboola (Brainforge)

Summary

Data Infrastructure Overview & Reliability

  • Stitch connectors generally stable
    • Google Sheets most error-prone due to manual data entry
    • Shopify and Ads Manager connectors reliable
  • Prefect deployment solid overall
    • Business Central API pipelines only consistent problem area
      • Frequent timeouts, both full refresh and incremental queries
      • Retry improvements may help but won’t fully solve
      • Alternative: Work with Lid team for custom endpoints
    • All syncs clearly named for easy identification
    • No memory or infrastructure issues
  • Token/credential management
    • Amazon updates keys every 6 months to 1 year
    • Environment file with all credentials coming via 1Password note to Mary
    • No frequent rotation requirements (nothing every 30 days)

DBT Models & Data Architecture

  • Core failure points minimal
    • Only fails on bad Google Sheets inputs or dev branch references
    • YAML config issues occasionally but easily fixed
  • Shopify data structure
    • Historical Canadian storefront unioned in (4+ years old data)
    • Current focus on US storefront
    • Custom “units” logic explodes line items into individual components
      • Multiple fulfillment methodologies handled over time
      • Component categorizations managed via Google Sheets
      • Well-documented but complex due to historical changes
  • Revenue model recently simplified
    • Now gross, net, and discount amounts
    • Unit-based revenue approach (non-standard)
    • Returns handling built into revenue model
  • Amazon integration straightforward
    • Orders become line items, same mapping logic as Shopify applies
    • Fulfilled shipments contain customer address data
  • Legacy cleanup opportunities
    • Retail data, Parabola workflows, House data all deprioritized
    • Data science folder no longer relevant
    • Some snapshots unnecessary

Key Transitions & Technical Debt

  • Major revenue metrics transition in progress
    • Moving from existing to new revenue definitions
    • All pieces ready, needs strategic execution
  • Business Central API migration
    • V2 endpoints built parallel to old ones
    • Transition timing depends on team comfort level
  • Shopify API version updates
    • Recent change (4-6 months ago) required ID parsing adjustments
    • Full historical backfill completed for US storefront
    • Canadian storefront still requires old joins
  • Klaviyo reverse ETL
    • Only reverse ETL process running
    • Three Prefect syncs, very stable
    • Profile updates infrequent
  • Redshift performance adequate
    • Most models don’t need incremental processing
    • No capacity issues
    • Some incrementality on Klaviyo events and BigQuery events only

Next Steps

  • Mary: Provide DBT admin permissions to Brainforge team
  • Roland: Send environment file with credentials via 1Password note to Mary
  • Brainforge: Set up email alerting for DBT failures
  • Brainforge: Consider Metaplane alert cleanup and customization
  • Roland available for follow-up questions (include Mary for coordination)