So I have two datasets and I have to match the value based on account number, currency, and amount but the tolerance should be 10 cent. Example:
Dataset 1:
Account | currency | amount |
1234 | USD | 10.20 |
1238 | USD | 10.10 |
2671 | EUR | 45.20 |
Dataset 2:
Account | currency | amount |
1234 | USD | 10.30 |
1234 | USD | 10.00 |
2671 | EUR | 45.20 |
1238 | USD | 10.40 |
Output:
Account | currency | amount | Result |
1234 | USD | 10.20 | Present |
1238 | USD | 10.10 | Not present |
2671 | EUR | 45.20 | Present |
1st row is present because the difference is +/- 10 cents and in dataset 2 the value is 10.30 (1st row), 2nd row is not present because 10.10 is not matching with 10.40 because our threshold is +-10 cents So 2nd line is not present
Hi @Sshasnk ,
You can use a join tool to bring the two datasets together and join on the fields [Account] and [Currency].
Note that out of the J output anchor you will get two rows for account 1234 with USD, one for the amount of 10.3 and one for 10 because your dataset 2 has 4 rows.
Then with a formula tool you can compare the two amount and create the difference.
Note that in your formula tool, you have to use a round function to round each amount separately. That's because of how computers store numbers, as they might appear as 10.2 but in reality they might be stored as 10.1999999999999 and when you compute the difference that might cause you some discrepancies.
Hope that helps, let me know if that worked for you.
Regards,
Angelos
@Sshasnk
Hope this helps
I kinda agree with @AngelosPachis and @Qiu
When doing work like this (reconciliations and partial-matches) - it's always important to look for duplicate matches.
For example - if data set 1 has USD: 10.00; and dataset 2 has two of these - then you need to spot that you are missing a match on the second one. This happens often when transactions have a common value.
User | Count |
---|---|
19 | |
15 | |
13 | |
9 | |
8 |