User Context & Persona

• Professional Profile

  • The user is a Data Analytics Engineer and Educator. They maintain a technical site at lelouvincx.com.
  • They are highly technical, preferring detailed, engineering-focused discussions over high-level overviews.
  • They are working at holistics.io

• Technical Stack & Expertise

  • Data Engineering: Expert in SQL optimization, data warehousing (Kimball/Inmon), and ETL/ELT pipelines. Frequently uses dbt, Dagster, DuckDB, PostgreSQL (RDS), and Redshift.
  • BI & Visualization: Experienced with Holistics, Metabase.
  • Infrastructure: Proponent of self-hosting and cost-optimization. Uses low-cost VPS providers (Contabo), Docker, Cloudflare.
  • Workflow: Uses MacOS with a keyboard-driven workflow. Preferred editor is Neovim (LazyVim config) and uses Logseq for knowledge management.

• Active Projects (Current Context)

  • 1. [Duty] Holistics Support

    • Role: Customer Duty Support / Data Analytics Engineer at Holistics.
    • Key Context: Focuses on troubleshooting and scaling solutions for clients.
    • Recent Work:
      • Client “Pencil”: Scaling customer dashboards and optimizing data models within the Holistics platform.
      • Workflow: Synthesizing support call transcripts and drafting technical follow-up communications.

  • 2. [Add2Cart] Data Consultancy Service

    • Tech Stack: RDS (PostgreSQL), S3 (data lake), Redshift (data warehouse), Holistics (BI tool).
    • Current Status: configuring replication tasks and debugging connection/permission issues between the OLTP source and OLAP destination.

  • 3. [Teaching] Data Analytics Educator

    • Role: Instructor for a junior analyst and a junior engineer.
    • Curriculum: Developing a gamified analytics course based on the Olist E-commerce dataset.
    • Resources: Creating a “Teacher’s Guide” and phased learning materials.
    • Infrastructure: Manages a self-hosted Metabase instance for students, powered by DuckDB and Neon.tech, exposed via Cloudflare Tunnels. Expose learning materials on public website at https://ngocyen99tb.lelouvincx.com, written in Docusaurus.

  • 4. [Personal] Infrastructure & Tools

    • Blog/Docs: Revamping lelouvincx.com (formerly hugo blowfish) (focus on themes, search plugins, and sitemaps).
    • Self-Hosting: Personal infrastructure includes Contabo VPS

• Cultural Context

  • The user is interested in Vietnamese language, culture, and technology (e.g., local banking, esports).

• https://gemini.google.com/app/87b8d4ee12216e39
• TODO Conceptual Design Priority: #P3
• TODO Techstack Design (after conceptual designing) Priority: #P3

• Category: [[Database Optimization]]
• When filter some data, instead of loading all data into memory then filter, predicate pushdown allows filtering at the query execution layer/storage layer, in order to only load the filtered data into memory.
• Benefit:
  • Reduce I/O
  • Reduce Time
• How it works
  • Before (without predicate pushdown)
    • Read: The engine read 100% data from storage. logseq.order-list-type:: number
    • Transfer: The engine transfers all data over the network to the compute engine. logseq.order-list-type:: number
    • Filter: The compute engine loads data into memory and applies WHERE. logseq.order-list-type:: number
  • After
    • Read: The compute engine sends the query and the filter to the storage/metadata layer. logseq.order-list-type:: number
    • Filter: The storage/metadata layer uses statistics based on the WHERE to skip entire redundant blocks of data. logseq.order-list-type:: number
    • Transfer: Smaller data transfers over the network. logseq.order-list-type:: number
• Pitfalls
  • Casting: something like WHERE CAST(string_col AS INTEGER) > 5 will not work.
    • Because the compute engine has to read all data to do the cast before filtering (which Holistics usually use).
  • Complex UDF.
  • Unsorted: If data random across row partitions, the predicate pushdown won’t work effectively.

• Principles

  • A triage page for heuristics gained from experience. Not the final home — the inbox / staging area.
  • Each entry: one-line trigger, one-line response, backlink to the journal where it was forged, and a status.
  • Once a heuristic matures (fires a 2nd time and has a clear domain), route it out to a dedicated page or upstream to a team-owned doc. The triage page stays small on purpose.

• How to use this page

  • Capture freely. Anything I notice once gets dropped here. Low friction.
  • Promote on second use. If a heuristic fires a 2nd time in real work, it has earned a spot on a dedicated page.
  • Archive on neglect. If an entry sits in triage > 60 days without a 2nd use, archive it. The page is for active heuristics, not aspirations.
  • Re-read during Friday wrap-up. Demote anything routed-out that hasn’t fired in 90 days.

• Routing targets

  • prioritization / scheduling → [[Priority Rules]]
  • sleep / illness / movement / energy → [[Health Rules]]
  • (future) work-craft pages may emerge:
    • communication → [[Communication Rules]] (TBD)
    • code review / on-call → [[Code Review Rules]] (TBD)
  • Some heuristics belong outside the second-brain. Upstream to team-owned docs when the rule is not unique to me — it converts a private heuristic into team infrastructure, and the writing exercise itself is career-visible.

• Status legend

  • triage — just captured, awaiting a 2nd occurrence.
  • routed → [[Page]] — promoted to a dedicated page; this entry kept as a stub for backlink.
  • upstream → <team doc> — proposed / merged into a team-owned doc.
  • archived — did not survive a 2nd use within 60 days, or superseded.

• Entries

  • Work craft

    • dbt deprecation 3-phase rule
      • trigger: deprecating a dbt model that other AML / BI assets reference.
      • response: phase 1 deprecate in code → phase 2 remove the .sql → phase 3 drop the BQ table. Don’t sit between phase 2 and 3 — AML references BQ physical names, so removing the .sql only stops refresh and leaves the table silently misleading consumers.
      • forged: [[Weekly Report - 2026-W17]] · Ampcode T-019da543
      • validated: W18 Calendly cleanup — 3 legacy models deleted in PR internal-aml-project#78 without re-deriving the reasoning.
      • status: upstream → holistics/dbt contributing guide (proposed). Team correctness rule, should outlive me.
    • Small P1 ships before big P2 context-switch
      • trigger: a P1 escalation lands while a small / cheap P1 task is still open.
      • response: ship the 30-min item first, then context-switch. The cheapest item is the one most at risk of slipping when escalations arrive.
      • forged: [[Weekly Report - 2026-W18]] (dbt PR #858 round 1 review slipped when BuyCo + Wamly P1 escalations landed Wed).
      • status: triage. Candidate for a future on-call / escalation handling note.

  • Communication

    • Customer comms reframe — recipe, not IOU
      • trigger: a customer asks for help with a workflow / configuration.
      • response: frame as “here is the recipe, happy to pair if helpful” instead of “I’ll do it for you”. Volunteering hands-on work tends to get silence in return.
      • forged: [[2026-04-27]] · validated against W17 Showbie + Basata silence pattern.
      • validated: W18 BuyCo 14-thread triage — replies were structured as workarounds + screen-recording asks, not “I’ll log in and fix it for you”.
      • status: upstream → presales comms playbook / CS onboarding doc (proposed). Pattern is not unique to my threads.
    • Ask for no, don’t ask for yes (Mooreds)
      • trigger: needing a small decision from a busy stakeholder (manager, peer).
      • response: replace “can we do X?” with “I’m going to do X to solve Y, will take care of it Monday unless I hear differently.” Shifts cognitive load to opt-out and unblocks progress while keeping ownership.
      • forged: [[Weekly Report - 2026-W17]] · mooreds.com
      • status: triage. Tone needs to be calibrated per person — practice on smaller decisions first.
    • Concision — what is the most important thing to say? (Wes Kao)
      • trigger: writing a status update, PR description, Slack message, or stakeholder note.
      • response: lead with the most important sentence. Cut anything that doesn’t change the reader’s decision or action.
      • forged: [[Weekly Report - 2026-W18]] · Wes Kao — How to Be Concise
      • status: triage. Pair with Levels of detail in communication once read.

  • Decision-making / scheduling

    • Vague task → time-block + decompose to smallest concrete next step
      • trigger: a task feels blocking / overwhelming and keeps getting deferred.
      • response: time-block a slot, then force decomposition into the smallest concrete next step. Long carry-overs often have a procrastination root, not a priority root.
      • forged: [[2026-04-30]] · Wait But Why — How to Beat Procrastination
      • validated: same-day BuyCo 14-item Notion page → 14 micro Slack threads in #holistics-buyco-external, all shipped in one Wed afternoon.
      • status: triage. Candidate for promotion to [[Priority Rules]] (pairs with the 3+ week carry-over forcing function) on next occurrence.
    • Highest-leverage triage on low-energy weeks → routed → [[Priority Rules]]
    • 3+ week carry-over forcing function → routed → [[Priority Rules]]

  • Reflection / craft hygiene

    • Writing is thinking — daily journals are 100% me
      • trigger: temptation to outsource a journal entry / reflection draft to an LLM.
      • response: don’t. LLM is fair game for boilerplate / first drafts only. The reflection muscle is the differentiator.
      • forged: [[Weekly Report - 2026-W18]] · HN thread
      • status: triage.
    • Cultivate non-LLM reading sources
      • trigger: defaulting to LLM search inside Logseq for ideas / craft inspiration.
      • response: deliberately seek out blogs / RSS / newsletters from individual senior writers (e.g., mooreds.com, Wes Kao, benn.substack). Signal/noise from a curated human writer is different from an aggregated answer.
      • forged: [[Weekly Report - 2026-W17]]
      • status: triage.

• Routed (kept as stubs for backlink)

  • Sleep > productivity → routed → [[Health Rules]]
  • Illness = hard stop on day 2 → routed → [[Health Rules]]
  • No gaming after 18:00 on weeknights → routed → [[Health Rules]]
  • Eat before exercise → routed → [[Health Rules]]
  • Weekend discipline (1 full off-day default) → routed → [[Health Rules]]

• Notes

  • Re-read this page during Friday wrap-up alongside [[Priority Rules]] and [[Health Rules]].
  • When the same heuristic shows up in 2 separate weekly reports, that’s the promotion signal — move it out.
  • Upstreaming to team docs is higher leverage than keeping a heuristic personal. Ask: “would Thuan / a new hire benefit from this being in the team’s docs?” If yes, draft the upstream PR.

• Priority Rules

  • A living set of personal rules for prioritizing tasks. Improve incrementally.

• Working principles

  • Choose the highest-leverage tasks even when the week’s workload is low (e.g., sick, leave, traveling). A low-energy week shouldn’t slide into a low-leverage week — pick fewer items, but pick the most important ones.
  • 3+ week carry-over rule — any item carried over 3 or more weeks gets a forcing function:
    • either elevate its priority and schedule it on a specific day, OR
    • formally move it to backlog / waiting (with the reason).
    • Goal: avoid the “always-on but never-done” zone.

• Priority tiers

  • P0 — Urgent
    • Must do immediately.
    • Usually customer-facing tasks that need a response or mitigation ASAP (e.g., production incident, enterprise customer escalation, data outage).
    • Drop other work to handle these.
  • P1 — High
    • Should prioritize doing because some stakeholders are blocking.
    • Includes manager-elevated items and teammate-blocking reviews.
  • P2 — Normal
    • Important but not urgent.
    • Picked up when no P1 tasks are active.
    • Most internal data work / improvements live here.
  • P3 — Personal / Teaching / Free time
    • Personal projects, teaching, learning, exploration.
    • Can be done in free time, or delegated to others when possible.

• Notes

  • Re-read this page during weekly planning (Friday wrap-up).
  • When a rule causes friction in practice, capture the friction here as a new bullet, then iterate.

Resources

• 1. The Foundational “Bible” of PLG
  • ProductLed.org: Run by Wes Bush. This is the gold standard for understanding the why.
  • Amplitude’s Mastering Retention & Growth Guides: While they sell a product, their playbooks on “Behavioral Cohorting” are essential for learning how to model user actions into “States” (e.g., Active, At-Risk, Power User).
  • Book - Product-Led Growth
• 2. Technical Modeling
  • dbt Labs Blog - Modeling for PLG: Search for their articles on “Modeling the Customer Journey.” They have excellent deep-dives on how to handle the many-to-many relationship between Users and Accounts in SQL.
  • The Hightouch or Census Blogs (Data Activation): These companies pioneered Reverse ETL. Their blogs are specifically geared toward engineers building the “HubSpot + Product Data” loop. Look for “How to build a PQL scoring model in Snowflake/BigQuery”.
  • Hacking Data (Newsletter/Blog): Often covers the technical debt involved in stitching together SaaS tools like HubSpot and Stripe (similar to your Zoho setup).
• 3. Industry Voices
  • Benn Stancil’s Substack: Benn (founder of Mode) writes deeply about the intersection of BI tools and data modeling. Since you are building a BI-related product, his insights on how users “consume” data are highly relevant.
  • Locally Optimistic Slack: A community for data leaders and engineers. There is often a lot of discussion about “GTM (Go-To-Market) Data Modeling.”

List of projects

• [[Add2Cart]] — Cross-functional project for the Add2Cart professional service: data pipeline, performance optimization, dynamic markdown, and S3 setup
• [[AI]] — Holistics AI squad work: test suites, AI feature trials, squad coordination with Dat/Tram
• [[Docs]] — Holistics documentation improvements and revamping
• [[Duty Support]] — Rotating on-call data support for Holistics customers (Zendesk tickets, impersonation, troubleshooting)
• [[Freetime]] — Low-priority reading and exploration done in spare time
• [[Internal]] — Holistics internal data team work: dbt, MRR reporting, CI, tracking
• [[Logseq]] — Improvements to this second-brain system: automation, backlog structure, weekly reports
• [[Misc]] — One-off tasks that don’t fit other projects
• [[Personal]] — Personal tooling, side projects, migration, self-hosting, learning
• [[Personal Productivity Analysis]] — Personal productivity monitoring: ingest weekly reports + journals into DuckDB (consistent schema with [[Personal Finance]] for later consolidation), build dashboard to track throughput, carry-overs, project mix, priority hit-rate, presales W/L, tooling ROI — drive data-driven planning and 1-on-1 decisions
• [[Presales]] — Holistics presales support: customer demos, solution proposals, competitive analysis
• [[Read]] — Reading and studying: articles, books, courses, youtube
• [[Smartclass]] — Personal side project: an ed-tech app with frontend (shadcn/ui) and releases
• [[Teaching]] — Mentoring and teaching data/SQL: composing projects (Olist), visualization, set-based thinking

• Source: https://www.holistics.io/blog/self-service-analytics-comic/
• Comments (Reddit Link)
  • 

• https://gemini.google.com/u/0/app/d7c56a75abd0fa3d