Rounding and Precision

Computers often struggle with decimal numbers (e.g., 0.1 + 0.2 might equal 0.30000000000000004). Beancount avoids this by using Decimal types, not floating-point numbers.

Precision Inference

Beancount does not have a hardcoded list of currencies. Instead, it infers the precision from your file.

If you write:

2024-01-01 * "Test"
  Assets:Cash   10.500 USD

Beancount will assume USD has 3 decimal places for the entire file. Best Practice: Always be consistent. Use 10.00 USD, not 10 USD.

The Tolerance Problem

In real life, splitting bills or currency conversion often results in fractions of a cent.

Example: Splitting $10.00 three ways.

• Person A: 3.33
• Person B: 3.33
• Person C: 3.33
• Total: 9.99

The transaction is off by 0.01.

Automatic Tolerance

Beancount allows a small tolerance (default 0.005) for transactions to balance. If the difference is within this tolerance, Beancount accepts it.

Accumulating Rounding Errors

If you want to track these discrepancies explicitly (or if they are larger than the tolerance), use the account_rounding option.

option "account_rounding" "Equity:Rounding"

When this is enabled, Beancount will automatically add a posting to Equity:Rounding to make the transaction balance perfectly.

2024-01-01 * "Split Bill"
  Expenses:Food       10.00 USD
  Assets:A           -3.33 USD
  Assets:B           -3.33 USD
  Assets:C           -3.33 USD
  ; Implicitly adds:
  ; Equity:Rounding  -0.01 USD