< blog
13 min read

Which Tableau or Looker Dashboards Drive Your Snowflake Bill? (2026)

Illustration of three dashboard screens on a stepped podium, the tallest coral screen drawing gold coins from a server block, showing which dashboard drives the Snowflake bill

You can rank Tableau dashboards by Snowflake credits, but Snowflake’s data alone cannot rank Looker dashboards. Tableau puts a workbook, worksheet or dashboard ID in the session tag of most queries, so you sum attributed credits by the ID inside that tag. Looker adds a comment with a user ID and a query run ID, not a dashboard. So you rank Looker’s repeated SQL and trace it back inside Looker. Every fact and figure in this guide was checked against Snowflake’s documentation.

This article is part of our guide to Snowflake cost monitoring. We checked Tableau and Looker behavior against their own help pages. For what each tool costs as a whole, see what Airflow, Fivetran and your BI tools cost in Snowflake.

What identifies a dashboard in Snowflake’s query records?

Only what the BI tool adds to each query. Tableau sets a session query tag with asset IDs, which Snowflake stores in the QUERY_TAG column. Looker appends a SQL comment, which can only appear in the query text. Neither tool sends a dashboard name, and only Tableau sends an ID at the dashboard level.

Snowflake’s QUERY_ATTRIBUTION_HISTORY view gives the compute credits of most queries, skipping those of about 100 milliseconds or less and Adaptive Warehouses. The SQL text lives in the QUERY_HISTORY view, which cuts QUERY_TEXT at 100K characters.

Tableau Looker
Added by default Session query tag SQL comment at the end
Snowflake column QUERY_TAG QUERY_TEXT
What it holds Asset, site and user IDs; user name on Server User ID, history slug, instance
Names the dashboard No, IDs only No
Rank by dashboard in Snowflake Partly, after parsing the tag No
Where to find names Tableau REST API Looker History Explore

What does Tableau’s query tag contain?

Tableau tags Snowflake queries by default on Tableau Cloud and on Tableau Server 2021.2 and later. The tag carries the workbook, worksheet or dashboard LUID, the site LUID and, when someone is signed in, the user LUID. The tag holds IDs, not workbook or dashboard names, and Tableau does not publish its exact format.

Tableau’s query tagging page lists the gaps:

  • Some queries may carry no asset ID. Ask Data, accelerated workbooks and Tableau Prep “are not currently supported”.
  • Backgrounder jobs carry no user.
  • Metadata queries are tagged too. On Tableau Server 2023.3 and later, expect small extra groups.
  • Server administrators can switch it off with a TSM setting.

Because the tag includes the user, one dashboard shows up under many tags. So group by the asset ID inside the tag. That ID may belong to a worksheet rather than its dashboard.

What does Looker add, and why can’t Snowflake name the dashboard?

Looker appends a context comment to the end of each query on Snowflake. It holds the Looker user ID, a history slug that identifies the query run, and the instance. It holds no dashboard, Look or Explore ID. So Snowflake’s data alone cannot rank Looker dashboards. Instead, you rank Looker’s repeated SQL.

Looker’s usage documentation lists Explores, SQL Runner, the API and filter suggestions as the comment’s sources, not dashboard tiles, so check one dashboard’s queries. It documents no field that joins a history slug to a dashboard.

Looker’s Snowflake connection page accepts QUERY_TAG as a JDBC parameter, but that labels the whole connection. No Snowflake page confirms that QUERY_TEXT keeps Looker’s comment, so check one query. A statement over 100K characters can lose it.

How do you rank dashboards by Snowflake credits?

Join QUERY_ATTRIBUTION_HISTORY to QUERY_HISTORY on QUERY_ID. Label each Tableau query by the asset ID inside its tag, and each Looker query by its parameterized hash. Then sum attributed compute and Query Acceleration credits by label. Look at a sample of Tableau tags first, because Tableau does not document their format.

Step 1: look at your Tableau tags

-- The 20 most common query tags for the Tableau user, last 7 days. Lags up to 8 hours. Needs SNOWFLAKE.USAGE_VIEWER or GOVERNANCE_VIEWER.
SELECT query_tag,
       COUNT(*) AS queries
FROM SNOWFLAKE.ACCOUNT_USAGE.QUERY_ATTRIBUTION_HISTORY
WHERE user_name = 'TABLEAU_SVC'
  AND start_time >= DATEADD(day, -7, CURRENT_TIMESTAMP())
GROUP BY query_tag
ORDER BY queries DESC
LIMIT 20;

Find the part of the tag that holds the asset ID. If the tags are JSON, TRY_PARSE_JSON returns NULL instead of an error on any tag that is not, and GET_PATH extracts the field at a path you name.

Step 2: rank the groups

A Looker group is one parameterized hash, which ignores comments and literals compared with =, !=, >= or <=, per Snowflake’s query hash guide. So a tile’s runs can group together despite each new history slug, while IN, LIKE, < or > filters can split them.

-- Top 25 BI query groups by attributed compute and QAS credits, last 7 days. QUERY_ATTRIBUTION_HISTORY lags up to 8 hours. Needs SNOWFLAKE.GOVERNANCE_VIEWER, which covers both views.
SELECT
  CASE
    WHEN qh.query_text ILIKE '%Looker Query Context%'
      THEN 'looker sql ' || qa.query_parameterized_hash
    WHEN GET_PATH(TRY_PARSE_JSON(qa.query_tag), 'replace.with.path') IS NOT NULL
      THEN 'tableau ' || GET_PATH(TRY_PARSE_JSON(qa.query_tag), 'replace.with.path')::STRING
    ELSE 'other ' || qa.user_name || ' ' || qa.query_parameterized_hash
  END                                                             AS dashboard_key,
  ROUND(SUM(qa.credits_attributed_compute
            + COALESCE(qa.credits_used_query_acceleration, 0)), 2) AS credits,
  COUNT(*)                                                        AS queries,
  COUNT(DISTINCT qa.user_name)                                    AS users,
  COUNT(DISTINCT qa.warehouse_name)                               AS warehouses,
  MAX(qa.query_id)                                                AS sample_query_id
FROM SNOWFLAKE.ACCOUNT_USAGE.QUERY_ATTRIBUTION_HISTORY qa
JOIN SNOWFLAKE.ACCOUNT_USAGE.QUERY_HISTORY qh
  ON qh.query_id = qa.query_id
WHERE qa.start_time >= DATEADD(day, -7, CURRENT_TIMESTAMP())
  AND qh.start_time >= DATEADD(day, -8, CURRENT_TIMESTAMP())
  AND (qa.user_name = 'TABLEAU_SVC'
       OR qh.query_text ILIKE '%Looker Query Context%')
GROUP BY 1
ORDER BY credits DESC
LIMIT 25;

Change the user name and the Step 1 path to your own. If your tags are not JSON, group by the whole tag and read each row as a dashboard and viewer pair. With Looker OAuth, each person runs as their own Snowflake user, so the query finds Looker by its comment. On Enterprise Edition or higher, Query Acceleration is on by default for new standard warehouses, so the query adds its credits.

GOVERNANCE_VIEWER reads both views, per Snowflake’s Account Usage roles. It covers 7 days because filtering QUERY_TEXT scans a lot of text. For tag basics, see our Snowflake query tags guide.

How do you turn the IDs into dashboard names?

Look them up in the BI tool. As an administrator, Tableau’s REST API lists each view with its ID, name, workbook and, optionally, usage statistics. Looker’s History Explore records each query’s user, time, source, runtime and dashboard ID, so match a costly query by Looker user ID and start time.

For Tableau, the workbooks and views methods include Get View, which returns the view’s workbook. Tableau’s IDs are locally unique IDs (LUIDs). Match the LUIDs from your tags against them, then roll worksheets up to their workbook.

For Looker, take a sample_query_id from the ranking and read its QUERY_TEXT and START_TIME in QUERY_HISTORY. The comment gives the Looker user_id. In the History Explore, filter on that user and a few minutes around that time, in the same time zone, and read the Dashboard ID. With no direct key, expect a few candidates. History deletes scheduled task rows after 7 days.

How should you read the ranking?

Read it as a floor, not a bill. Attributed credits leave out warehouse idle time, queries of about 100 milliseconds or less, cloud services and serverless features. So a dashboard that refreshes all day can show a small figure while keeping its warehouse running.

Refresh schedules and idle time

Warehouses bill per second with a 60-second minimum each time they start or resume. CREATE WAREHOUSE sets AUTO_SUSPEND to 600 seconds by default. In this illustrative case, one dashboard refreshes every 15 minutes, around the clock, alone on an X-Small Gen2 warehouse on AWS, with 30 seconds of queries each time.

Illustrative, 30 days Credits Enterprise, AWS US East, on demand
Billed, AUTO_SUSPEND 600 680.4 $2,041.20
Billed, AUTO_SUSPEND 60 97.2 $291.60
Attributed to queries 32.4 $97.20

New standard warehouses default to Gen2, which bills 1.35 credits an hour at X-Small on AWS. The queries account for 4.8% of what the warehouse bills at the default setting. The billed rows are a floor, because Snowflake checks for suspension about every 30 seconds, and they leave out Query Acceleration. To see the gap on your own BI warehouses:

-- Idle credits per warehouse, last 30 full days: metered compute minus credits attributed to queries. Lags up to 3 hours. Needs SNOWFLAKE.USAGE_VIEWER.
SELECT warehouse_name,
       ROUND(SUM(credits_used_compute), 2) AS compute_credits,
       ROUND(SUM(credits_used_compute) - SUM(credits_attributed_compute_queries), 2) AS idle_credits
FROM SNOWFLAKE.ACCOUNT_USAGE.WAREHOUSE_METERING_HISTORY
WHERE start_time >= DATEADD(day, -30, CURRENT_DATE())
  AND end_time < CURRENT_DATE()
  AND warehouse_name IN ('BI_WH')
GROUP BY warehouse_name;

This follows the WAREHOUSE_METERING_HISTORY idle-time example. To share idle credits across users, see Snowflake cost by user, role and warehouse.

Extracts and caches

A Tableau extract is a subset of data saved apart from the source and refreshed from it. So Snowflake sees the refresh, not each person who opens the dashboard. On Tableau Cloud, refresh schedules can run hourly. On Tableau Server, the Backgrounder runs them. Tableau does not say which asset ID a refresh carries, so expect some refresh cost in the other rows.

Snowflake can reuse a persisted query result for 24 hours when the query text matches exactly and the data is unchanged, and reuse bypasses query execution. Snowflake does not mention comments, so assume Looker runs rarely reuse it. Looker’s own cache keeps results for one hour by default.

What should you do about the top few dashboards?

Check each one’s views against its credits, then fix the schedule before the SQL. Retire or slow down dashboards that cost a lot and are rarely opened. Tie refreshes and caches to when the data changes. Then tune the few queries that stay on top.

  1. Compare cost with use. Tableau’s view usage statistics and Looker’s Content Usage Explore show what people open.
  2. Match refresh to data. Refresh a Tableau extract when its source loads. In Looker, a datagroup ties cache expiry and derived table rebuilds to your ETL.
  3. Right-size idle time. Set auto-suspend to fit the gap between refreshes. Our Snowflake auto-suspend guide covers the trade-off: Snowflake drops the warehouse cache when a warehouse suspends.
  4. Tune what is left. Check the sample query’s bytes scanned and partitions scanned in QUERY_HISTORY.
  5. Re-run the ranking after each change.

How Seemore helps with dashboard cost

The routine above works, but someone has to run it and read the results. The Seemore platform reads the query tag Tableau adds to most of its queries and the comment Looker appends, so you can group warehouse cost by Tableau workbook or dashboard, or by Looker user. Seemore’s Cost Anomaly Detection compares each day’s spend for the accounts and warehouses its rules cover against the previous 14 days and explains the likely root cause of each spike or drop. You can also ask Seemore Assistant about your environment in plain language instead of writing the SQL above.

Frequently asked questions

How do I find the most expensive Tableau dashboards in Snowflake?

Sum CREDITS_ATTRIBUTED_COMPUTE in QUERY_ATTRIBUTION_HISTORY by the workbook, worksheet or dashboard ID inside Tableau’s session query tag, not the whole tag, which also holds the user. Look up names with Tableau’s REST API. Some features and background jobs carry no asset ID or user.

Does Looker set a query tag in Snowflake?

Not by default. Looker appends a comment to the end of the SQL instead, with a user ID and a history slug. You can add QUERY_TAG to a Snowflake connection’s additional JDBC parameters, but that labels every query on the connection, not a single dashboard.

Do Tableau extracts use Snowflake credits?

Yes, when they refresh. An extract is a copy saved apart from Snowflake, so Snowflake sees the refresh queries rather than each dashboard view. Scheduled refreshes can run as often as hourly on Tableau Cloud, and Tableau Server runs them as backgrounder jobs.

Does caching lower Snowflake dashboard cost?

It can. Snowflake can reuse a persisted result for 24 hours when the query text matches exactly and the data is unchanged. Looker also caches results for one hour by default, before any query reaches Snowflake. Cached runs largely drop out of the ranking, so a busy dashboard can rank below a rarely opened one.

How do I find unused dashboards that still add to Snowflake dashboard cost?

Compare the credit ranking with each tool’s own usage data. Tableau’s REST API can return usage statistics for every view, and Looker’s Content Usage Explore shows which content has not been accessed in the past 60 days.

How often does Looker keep a Snowflake warehouse awake?

With persistent derived tables enabled, Looker checks for regeneration every five minutes by default, and Looker says this keeps Snowflake warehouses active. Change the regeneration schedule if warehouses should suspend overnight.