Omni vs Tableau — What you’re actually looking at
Audience: All hands · Purpose: Same business data, different “when” and “how” it reaches your screen.
The short version
Omni and Tableau both read from Eden’s data warehouse (BigQuery, built by dbt). Neither tool creates the numbers first—they reflect whatever finished loading in the warehouse last.
If a chart in Omni ever looks different from what you remember in Tableau, it is usually one of: when the snapshot was taken, how the date range is defined (for example, whether “today” counts), or where a metric is defined—not a mystery bug.
Where freshness really comes from
| Layer | What it does |
|---|---|
| dbt / pipelines | Refreshes tables and marts on a schedule. This is the main “heartbeat” for how current core reporting is. |
| Tableau | Often showed a daily snapshot (extract refreshed on a fixed schedule). You could be seeing data as of that run, not the latest second in the warehouse. |
| Omni | Queries the warehouse live through curated Topics. For speed, dashboards may use caching; refreshing or re-running can show the latest query result. |
Takeaway: Day-to-day “how fresh is this metric?” is mostly about pipeline timing, not Omni vs Tableau as brands.
Why two tools might show different numbers for the “same” thing
-
Snapshot vs live — Tableau extracts are once per day; Omni aligns with what BigQuery returns now (after the last successful pipeline run). Comparing them at different times of day can disagree.
-
Date filters — Phrases like “last 30 days” can include or exclude today, or differ by a day at the boundaries. Small wording differences change which rows are in the chart.
-
How metrics are built — Some logic lived in Tableau calculations. In Omni, many definitions live in the model / Topic so everyone uses the same rule. The principle is consistency; the implementation is different.
-
Email and Slack — Scheduled reports (morning emails, Slack images) send a picture of a moment. That moment might not match someone who opens the live dashboard later the same day.
What to do if something looks off
- Check when the warehouse last updated (if your team publishes that) and when you’re comparing tools or screenshots.
- Prefer one definition for leadership decisions (same dashboard, same filters)—and use scheduled deliveries if you need a repeatable daily snapshot.
- Tell us when something doesn’t match your expectations—we use that to tighten the product.
How we keep improving (Blobby, definitions, transparency)
As real questions come in, we can refine how Blobby (Omni’s AI assistant) answers—better context, clearer guardrails, and fewer ambiguous interpretations.
We can also set strict, documented definitions for phrases everyone uses loosely (for example “last 30 days”: inclusive or exclusive of today, calendar vs rolling, timezone). Same words should mean the same thing in dashboards and in AI.
Over time we can surface calculations more transparently so it’s obvious what a number includes and how it was built.
You’re encouraged to share feedback whenever something seems wrong, unclear, or harder than it should be. We’ll adjust the tool and the model so Omni stays accurate and useful for how Eden actually works.
One sentence for the room
Tableau often showed a scheduled daily snapshot; Omni reflects the warehouse after the latest pipeline run, with clear Topics and optional caching—so differences are usually timing, filters, or definitions, not randomness.
Internal reference: Eden Tableau → Omni migration QA and data platform standups. For tool training, see stakeholder training materials in this folder.