Four kinds of precision
Dans ce modèle pédagogique, A rate may be stored more precisely than displayed. Accrued income may retain fractions of a kopeck. A credited payment is usually shown as money, while total yield is rounded as a percentage. These are different presentation levels. Know what is rounded, when and by which rule.
An experiment with two algorithms
Dans ce modèle pédagogique, Assume a constant base of 12,345 roubles, a fictional simple annual rate of 7%, exactly 31 days and a 365-day denominator, with no capitalization or movements. Retaining fractional kopecks until the end gives 12,345 × 0.07 × 31 / 365 = 73.39 roubles after rounding. Rounding each identical daily accrual to 2.37 gives 73.47 over 31 days. The difference is 8 kopecks.
Dans ce modèle pédagogique, Both results are reproducible, but neither can be declared contractually correct without the terms. The example uses half-up rounding and decimal arithmetic. Daily capitalization creates a new experiment because the base changes; a leap year or different denominator also changes the model.
Accrual is not crediting
Dans ce modèle pédagogique, Check that like entities are compared. Informational daily accrual and an end-of-term payment may have different presentation precision. Verify start and end dates, top-up timing and balance base. An extra day often matters more than several kopecks. Start with a short no-movement period, then add capitalization and transfers.
Annualization adds another uncertainty
Dans ce modèle pédagogique, If 30 educational days produce 1%, repeating that result over a 365-day year gives (1.01)^(365/30) − 1, approximately 12.87%. This is not actual annual income or a forecast that the short result repeats. Simple proportion and effective repetition answer different questions, and an already rounded 1% makes the annualized figure approximate.
A process for investigating the difference
Dans ce modèle pédagogique, Keep original documents and the calculation version. Compare amount, dates, rates and units, day-count basis, event order, rounding point and deductions. Do not alter one parameter merely to match a statement. If the rule is unknown, state the unresolved difference and ask precisely about the method. Reconciliation aims at reproducible causes, not a preselected accusation or excuse.
Do not trust cell formatting automatically
Dans ce modèle pédagogique, Two decimal places may change only the display while retaining more internal digits; visibly rounded rows can therefore differ from a visible total. Apply explicit rounding in a separate column and compare methods without replacing source data. Binary arithmetic can approximate money; this control experiment uses decimal arithmetic. First establish the accrual method, then implement it carefully.