Detected columns with numeric type and unspecified…

Error message

Detected columns with numeric type and unspecified precision/scale, this can lead to unintended rounding: {col_naked_numeric}`

What it means

This warning is emitted when a model column resolves to a numeric (NUMERIC/DECIMAL) type without an explicit precision and scale. Without precision/scale, the database applies its default, which can silently round or truncate values. dbt raises it during get-empty-relation / get_column_schema_from_query work when casting columns via cast().

Solutions

Add explicit precision and scale to the column type, e.g. cast the column to numeric(38, 9) instead of numeric.
If using model contracts, set the data_type in the contract's columns block to numeric(p, s).
If the database's default numeric behavior is acceptable, acknowledge the warning and move on — it is non-fatal.

Example fix

// before
{{ cast('null', 'numeric') }} as my_col
// after
{{ cast('null', 'numeric(38, 9)') }} as my_col

Defensive patterns

Strategy: validation

Validate before calling

# dbt schema.yml / model contract check before running
models:
  - name: my_model
    columns:
      - name: my_col
        data_type: numeric(38, 9)  # specify precision & scale, not bare 'numeric'

Prevention

Always write numeric column types as numeric(p, s) in contracts and casts.
Add a pre-commit lint that flags bare numeric/decimal in dbt YAML files.
Verify inferred types by inspecting the built relation after first run.

When it happens

Trigger: A column's data_type is a bare numeric type (e.g. 'numeric', 'decimal') with no '(p,s)' specifier while dbt builds the empty relation or derives a schema from a query; the column passes the col_err check but lands in col_naked_numeric.

Common situations: Inferencing schemas fromCTAS/scratch relations where types come back as unadorned 'numeric'; writing user-defined column types in a model contract without precision; adapters whose default numeric precision differs from what the user expects.

Understand the failure class

Background: "Invalid ... format", "must be in format X", "does not look like a ..." — invalid argument format errors across CLI tools and libraries — this error's family across 17 libraries.

Related errors

AI-assisted analysis of dbt-labs/dbt-core@0267ce9170 (2026-09-07). Data as JSON: /api/errors/0a85a5fe9f255df0. Report an issue: GitHub.

Appendix: source

Thrown at crates/dbt-loader/src/dbt_macro_assets/dbt-adapters/macros/adapters/columns.sql:77

    {%- set col_err = [] -%}
    {%- set col_naked_numeric = [] -%}
    select
    {% for i in columns %}
      {%- set col = columns[i] -%}
      {%- if col['data_type'] is not defined -%}
        {%- do col_err.append(col['name']) -%}
      {#-- If this column's type is just 'numeric' then it is missing precision/scale, raise a warning --#}
      {#-- TYPE CHECK: col['data_type'] is optional[string] but user have this constraint --#}
      {%- elif col['data_type'].strip().lower() in ('numeric', 'decimal', 'number') -%}
        {%- do col_naked_numeric.append(col['name']) -%}
      {%- endif -%}
      {% set col_name = adapter.quote(col['name']) if col.get('quote') else col['name'] %}
      {{ cast('null', col['data_type']) }} as {{ col_name }}{{ ", " if not loop.last }}
    {%- endfor -%}
    {%- if (col_err | length) > 0 -%}
      {{ exceptions.column_type_missing(column_names=col_err) }}
    {%- elif (col_naked_numeric | length) > 0 -%}
      {{ exceptions.warn("Detected columns with numeric type and unspecified precision/scale, this can lead to unintended rounding: " ~ col_naked_numeric ~ "`") }}
    {%- endif -%}
{% endmacro %}

-- funcsign: (string, optional[string]) -> list[base_column]
{% macro get_column_schema_from_query(select_sql, select_sql_header=none) -%}
    {% set columns = [] %}
    {# -- Using an 'empty subquery' here to get the same schema as the given select_sql statement, without necessitating a data scan.#}
    {% set sql = get_empty_subquery_sql(select_sql, select_sql_header) %}
    {% set column_schema = adapter.get_column_schema_from_query(sql) %}
    {{ return(column_schema) }}
{% endmacro %}

-- here for back compat
-- funcsign: (string) -> list[string]
{% macro get_columns_in_query(select_sql) -%}
  {{ return(adapter.dispatch('get_columns_in_query', 'dbt')(select_sql)) }}
{% endmacro %}

View on GitHub (pinned to 0267ce9170)