Double payment analysis for non-SAP data
Introduction
A double payment analysis reveals invoices that have potentially been posted more than once.
If these postings are offset by payments at a later stage in the process, multiple payments are created for the same transaction. We use our standard AP_Dupl_Payments analysis to identify these multiple postings on SAP data, which shows incoming invoices that have potentially been posted multiple times.
Some booking transactions appear to be duplicates, but are actually correct. These are referred to as potential hits and require a detailed check. By labelling confirmed duplicate bookings, an AI model can be trained to automatically evaluate new cases and reliably identify false positives so that they can be excluded.
Customers who do not use SAP but would like to benefit from AI evaluation can use our AI_Dupl_Payments analysis tool. This requires specially identified potential duplicate bookings to be transferred in a predefined format.
Requirements
- Installing ACL for Windows & local installation dab AnalyticSuite: Analysis can only be performed locally on the desktop, not in Robotics
- R and R-Studio: Installation required, R.exe and R package required (article on installing the R environment)
- Highbond: Results module required
- TF_Import.xlsx: File must be stored in the same path as the ACL project
- Initial Pool file with the name AI_Dupl_Payments_Pool.xlsx:
- Name of the file must be AI_Dupl_Payments_Pool.xlsx
- Path (default C:\Users\Public\dabgmbh\AnalyticSuite\Utilities) can be changed
- File must contain duplicate payments that have already be found, checked and marked as True Duplicate Payments or False Duplicate Payments
- File with data that has not yet been verified: must be saved in the defined path
- Column list for the table: required in the Pool file and in the file with the unchecked data
- Difference between the two files is that the AI_Meta_Information, AI_Potential, AI_Assessment and Manual Assessment fields are NOT presend in the file with the unchecked data.
Settings in the start script __Run_AI_Dupl_Payments
The following variables are required for uploading to the Highbond Results module:
- v_TF_Cust_Specific_OrgID: Organisation-ID of the Entity
- v_TF_Cust_Specific_UploadToken: Highbond API token of the user performing the analysis
- v_TF_Cust_Specific_APIArea: Possible values: 1 = EU (Default); 2 = US; 3 = CA; 4 = AS; 5 = AU
- v_AI_Dupl_Paym_OneLineID:
- ID of the Highbond table where the results are uploaded
- The Highbond table is not created until the upload during the first run, which means that the variable must not exist at the start of the first run
- The corresponding line must be commented
- After the first Run: The Highbond address must be entered in the table to enable further updates of the Highbond table in subsequent runs
The following variables have default values that can be customised
- v_AI_Dupl_Paym_AI_ExpPath: Path where the pool file is stored and updated. Recommendation: Use the predefined path - as the pool file is updated after each run, the path must not be changed after the first run.
- v_AI_Dupl_Paym_Unseen_Path: Path and name of the unseen data file to import
- v_TF17_ExportPfad: Path where the .json file will be created
Procedure in the first run
- Pool file and the unchecked data file are processed in the R script
- AI_Meta_Information, AI_Potential, AI_Assessments fields created in the R script are transferred to the ACL script and added to the unchecked data
- Entries with a unique ID are duplicated: one row per unique ID in table AI_Dupl_Payments_OneLine
- AI_Dupl_Payments_OneLine table is uploaded to the Highbond Results module, where a new table is created
Manual steps to be taken after the first run in Highbond
- Create a questionnaire for the collection created
- Assign a name, e.g. detection of real or false duplicate payments
- Select and save uploaded table in Linked tables
- Create a new question (type: single choice)
- Question e.g. Is this record a true or false duplicate payment?
- Add two possible answers: True double payment, False double payment
- Important: Customise column names in manual scoring
- In the Highbond table settings, define the UpdateKey field as the primary key
- Process the records in the questionnaire: The Manual_Assessment_Upl field contains records from the initial pool file:
- Field is filled: Values can be accepted
- Remaining entries: User decision required based on the values in the AI_Meta_Information, AI_Potential and AI_Assessments fields. The Manual_Assessment_Upl field can be deleted once all values have been accepted.
Manual steps before the second run in the ACL start script __Run_AI_Dupl_Payments
- Copy the ID of the Highbond results table: Paste the value into the v_AI_Dupl_Paym_OneLineID variable. Required to import the Highbond result table into the ACL.
- TF_Import: Open the file and enter the ID in the ScheduledResultsUpload sheet. This is necessary to update the Highbond table with the new entries.
- File unseen_date: Refresh
Procedure in the following runs
- Highbond Table: is imported into the ACL as it contains manual user ratings
- Pool file: Updated with information from the Highbond table
- Pool file and file with the unchecked data: Processed in the R script
- AI_Meta_Information, AI_Potential, AI_Assessments: fields created in the R script are transferred to the ACL script and added to the unchecked data.
- Entries with a unique ID are combined: To obtain one row per unique ID in the AI_Dupl_Payments_OneLine table
- Table Dupl_Payments_OneLine: Uploaded to the Highbond Results module where the existing Results table is updated
