Move Tableau calculations into dbt models.

What is the direct answer to this workbook question?

Converting Tableau calculated fields to dbt models requires splitting workbook formulas into three layers: move row-level IF and string expressions into dbt staging models, refactor FIXED Level of Detail formulas into intermediate CTEs with explicit GROUP BY grains, and define ratio aggregations in your mart or semantic layer. Inspect the .twbx formulas first in the xmetric inspector.

Why should I separate calculation grains before dbt migration?

A Tableau workbook mixes three different calculation grains inside the same sidebar list: row-level expressions, view-dependent aggregations (SUM([Profit]) / SUM([Sales])), and grain-overriding Level of Detail ({ FIXED }) expressions.

If an analytics engineer or AI coding agent copies an aggregate ratio like SUM(profit) / SUM(sales) into a daily dbt mart table and later sums that column by month in a BI tool, the ratio of sums turns into a wrong sum of ratios. Classify every workbook formula by grain before writing your dbt SQL.

Which dbt layer fits each calculation type?

Keep row-level expressions in staging, explicit grouped calculations in intermediate models, and aggregate business metrics at their intended reporting grain.

How can I refactor a workbook calculation into dbt SQL?

Suppose you drop a legacy .twbx into the xmetric workbook inspector and find three nested calculated fields used on the executive revenue sheet:

// 1. [Clean Status] (Row-level)
IIF(ISNULL([Cancel Date]), 'Active', 'Churned')

// 2. [First Active Order] (FIXED LOD)
{ FIXED [Customer ID] : MIN(IF [Clean Status] = 'Active' THEN [Order Date] END) }

// 3. [Is New Customer Order] (Row-level referencing LOD)
[Order Date] = [First Active Order]

What belongs in the staging model?

Move the row-level status cleanup into the staging model so every downstream model inherits it:

with source as (
    select * from {{ source('raw', 'orders') }}
)

select
    order_id,
    customer_id,
    order_date,
    cancel_date,
    amount,
    case
        when cancel_date is null then 'Active'
        else 'Churned'
    end as clean_status
from source

What belongs in the intermediate or mart model?

Translate the FIXED LOD into a clean CTE and join it back on customer_id:

with orders as (
    select * from {{ ref('stg_orders') }}
),

customer_first_active as (
    select
        customer_id,
        min(case when clean_status = 'Active' then order_date end) as first_active_order_date
    from orders
    group by customer_id
)

select
    o.order_id,
    o.customer_id,
    o.order_date,
    o.clean_status,
    o.amount,
    c.first_active_order_date,
    (o.order_date = c.first_active_order_date) as is_new_customer_order
from orders o
left join customer_first_active c
    on o.customer_id = c.customer_id

How can I verify the dbt output against the workbook?

Never decommission a workbook until the new dbt table produces the exact same numbers as the original sheet:

1. Inspect the original .twbx in the xmetric workbook inspector in your browser to save the packaged data files and list the worksheet's formulas. You can also open the workbook in the xmetric app to rebuild the sheet on shelves.
2. Test the translated logic by running the original Tableau formula in the formula playground, which shows the equivalent SQL, or import your exported dbt output (CSV, Excel, or Parquet) into the xmetric app and explore it on shelves in your browser. Live Postgres and BigQuery connections remain coming.
3. Compare the aggregated totals across dimensions to confirm that filter order of operations and null handling match 100%.

What data does this worked example use?

The initial view uses authored synthetic data, not customer data or a captured agent session.

The example record lists its file, calculation or SQL and expected values. The live browser result is computed by xmetric’s existing engine or workbook inspector. Samples are excluded from own-data activation reporting.