Chinh (lelouvincx) / 2026-06-09

Created Tue, 09 Jun 2026 00:00:00 +0000 Modified Thu, 11 Jun 2026 05:26:03 +0000
310 Words
journal

Notes

  • https://www.reddit.com/r/BusinessIntelligence/comments/1u0ywdb/how_are_data_teams_letting_nonengineers_configure/

  • The gray area between dbt and BI, here is an example: reddit

  • Tasks

  • Partly today

  • Done

    • DONE Prepare for onboarding call 1 with [[dextech.ai]]

  • Problems:

    • Self serve capabilities - dashboard interactivity (drill through, drill down, break down, view underlying data), personal workspace, dataset explore, conversational AI
    • Superset to Holistics logseq.order-list-type:: number
    • Multi-environment setup logseq.order-list-type:: number
    • JSONB + performance on postgres (RDS) logseq.order-list-type:: number

  • Questions to clarify:

    • Are their end users almost non-tech folks, or they know how to use any coding tools?
    • Which 1–2 Superset dashboards have the highest customer usage? Those are the ones we should walk through migration on first.

  • I am preparing for the call, to recap his questions: (1) learn migration process of embedded dashboards from Superset to Holistics; (2) learn Holistics development workflow

  • I studied Superset concepts and find that the most different of superset compared to holistics is it does not have a concept dataset (which is a set of models, join config, row-level security). it may explain the reason they are looking for a new self-serve BI tool. or more important, changing their current waterfall reporting process

  • Their pain point is the current reporting process has a long development workflow/process from customer request -> result. they want a self serve solution

  • Our self serve solution mostly are dashboard interactivity, dataset exploration, conversational AI. all requires well designed, well governed datasets

  • Product-wise, superset lacks this

  • If holistics is going to replace superset just for dashboard embedding, it does not solve their pain point

  • My idea for today’s call is to present them this problem, and recommend direction of the POC: (1) start with 1-2 most used superset dashboards (2) replicate those dashboards 1:1 in holistics, means, heavy-model + 0 relationship dataset (to create a sense of possibility) (3) refactor, re-model datasets into holistics-native semantics

  • I have drafted a slide deck here: slack