Tableau Performance and Extracts: Speed Up Dashboards

Even brilliant visuals fail if a dashboard takes 20 seconds to filter. Built from the performance threads in script 01 (Live vs Extract, source filters) and script 11 (dashboard layout), this guide gives analysts the speed playbook — when Hyper wins, how source filters shrink cost, and how cadence decisions protect trust at ₹5–10 LPA expectations.

Connect this to Tableau Connecting Data Sources, structure via Tableau Dashboards, and narrate via Tableau Storytelling with Data. Educational only — platform selection lives in the future comparison capstone.

Why does performance start at the connection?

Every viz queries the Data Source layer Tableau Help: Extract. If the layer carries 50M unfiltered rows live from a CSV, every filter re-scans them. Fixing later with sheet filters is too late — the cost is already paid.

Two levers at the Data Source page (script 01 at 1:32):

• Live vs Extract toggle — decides where compute lives.
• Data source filters — decides how much data enters.

Get both right and dashboards pass the 5-second scan; ignore them and even Hyper struggles.

When do Hyper extracts beat Live?

Hyper = columnar, compressed, in-memory scan Tableau Help: Extract and Gartner 2024.

Use Live when a Snowflake or BigQuery warehouse already serves BI concurrency cheaply and stakeholders need minute-fresh ops. Use Hyper when the source is CSV/Excel, the DB is slow, or 20 viewers would hammer production. Hyper extracts are publication-ready artefacts for portfolio stories .

How do source filters cut cost?

Permanent row exclusion Tableau Help: Extract. At top-right Filters on the Data Source page:

// Example source filter (conceptual)
// Exclude test accounts and pre-2020 history before extract
// [Account Type] != "Test"  AND  YEAR([Order Date]) >= 2020

Why stronger than sheet filters:

• Sheet filter: queries all rows, then hides excluded marks (still scanned).
• Source filter: excluded rows never enter Hyper or the Live query — smaller extract, faster scans, lower DB cost.

Rule: promotional or QA rows that never answer a business question belong in a source filter, not a dashboard toggle.

How do you manage refresh and publish cadence?

Schedule like a data SLA Tableau Help: Performance.

1. Estimate freshness need: hourly for support, nightly for sales, weekly for static wildlife-strikes-style stories.
2. Schedule on Tableau Cloud/Server: set refresh time, notify on failure, validate row delta post-refresh.
3. Propagate to stories: story points are live-linked; an extract refresh updates stories without rebuild Tableau Help: Story Present.
4. Version: keep one published Hyper per environment; avoid desktop-local extracts that drift.

Quick checklist

• Data source filters applied for permanent exclusions
• Live vs Hyper chosen per freshness table
• Extract scheduled and timestamp shown on dashboard
• Dashboard layout tiled (see Tableau Dashboards) before speed test
• Story points validated post-refresh (see Storytelling)
• Next: practice on Data Analyst track

Treat performance as part of design, not an afterthought. A fast dashboard gets opened daily; a slow one gets replaced by a spreadsheet export.