SnowPro Associate: Platform Question 115
Single answer● Data typesA retail company is loading order data from CSV files into Snowflake. One column, ORDER_TOTAL, contains values such as 125.50, 89, and 1000000.99. Finance reports that some totals are being rounded unexpectedly after load. The current target table defines ORDER_TOTAL as FLOAT. The team wants to preserve exact values for reporting and downstream calculations. Which change should the Snowflake engineer make?
- A
Change ORDER_TOTAL to NUMBER(38,2) to store exact fixed-point decimal values
- B
Change ORDER_TOTAL to VARCHAR so the original text is preserved exactly for calculations
- C
Keep ORDER_TOTAL as FLOAT and increase the virtual warehouse size during loading
- D
Change ORDER_TOTAL to BOOLEAN because it uses fewer bytes and improves numeric precision
Show answer and explanation
Correct answer: A
Explanation
For currency and other exact numeric values, Snowflake best practice is to use a fixed-point type such as NUMBER, DECIMAL, or NUMERIC rather than FLOAT. FLOAT is an approximate numeric type and can introduce representation differences that appear as rounding issues in financial reporting or calculations. NUMBER(38,2) is a common design choice for monetary amounts because it supports exact precision with two decimal places. Snowflake documentation describes NUMBER/DECIMAL/NUMERIC as fixed-point numeric types and FLOAT/DOUBLE as approximate numeric types, making fixed-point the appropriate choice for finance-related columns.
- A. Correct.
Correct. In Snowflake, NUMBER/DECIMAL/NUMERIC are fixed-point numeric types and are appropriate for exact values such as currency. Defining ORDER_TOTAL as NUMBER(38,2) preserves two decimal places without the approximation behavior associated with floating-point storage. This is the recommended approach for financial amounts where exact arithmetic matters.
- B. Incorrect.
Incorrect. VARCHAR would preserve the textual representation, but it is not appropriate for calculations, aggregations, comparisons, or financial reporting logic without repeated casting. Storing currency as text introduces data quality and performance issues rather than solving the underlying type selection problem.
- C. Incorrect.
Incorrect. FLOAT is an approximate numeric data type. Increasing warehouse size can improve compute performance, but it does not change how values are represented or eliminate floating-point precision and rounding behavior. This option reflects the misconception that compute size affects data type accuracy.
- D. Incorrect.
Incorrect. BOOLEAN stores logical TRUE/FALSE values, not monetary amounts. It cannot represent decimal order totals. This distractor targets confusion between storage efficiency and suitability of a data type for the business requirement.