Please consider following example:
| Before Rounding: | | | |
| | Jan | Feb | Cum |
| Item 1 | 12.534145 | 8.126270 | 20.660415 |
| Item 2 | 14.004458 | 7.119500 | 21.123958 |
| Total | 26.538603 | 15.245770 | 41.784373 |
| | | | |
| Rounded Version | | | |
| | Jan | Feb | Cum |
| Item 1 | 12.53 | 8.13 | 20.66 |
| Plan 2 | 14.00 | 7.12 | 21.12 |
| Total | 26.54 | 15.25 | 41.78 |
There is no problem in calculation in "Before Rounding" version.
However, user may report that some problem on Rounded version. Because 26.54+15.25 <> 41.78
But, if we patch the data become 41.79, then it become another problem since 20.66+21.12<> 41.79
It is a perception dilemma because we did not recognized that all version are rounded values rather than actual values.
May be you may suggest to do rounding before all calculations, however, please aware that cumulative error will become very large, as we have more than 5,000 of item in calculation .
Let me compare the "error"
| Method | Error due to Rounding | Error due to Human Perception |
| Rounding only on display (Current method) | 0.00 (as no rounding in calculation) | 0.005 |
| Rounding before calculation | =0.005* 5,000 = 25.00 | 0.00 |
Conclusion:
The "Rounding Issues" is because of a wrong perception from human rather than system error.
