In the world of account reconciliation, the primary goal of Transaction Matching is straightforward: load your data, apply predefined rules to match transactions, and manage the exceptions that fall through. Automated rules handle the bulk of the work — but the remaining unmatched transactions can still demand significant manual effort from your accounting team.
That is where Transaction Matching Assistance enters the picture — leveraging machine learning algorithms to predict matches, reduce exceptions, and streamline the close process.
"Automated rules catch the majority. AI catches what rules miss. Together, they eliminate virtually all manual reconciliation toil."
How It Works
Transaction Matching Assistance analyzes patterns within your historical manual match data to train an intelligent prediction model. The workflow is designed to run seamlessly within your existing close process:
⚡ Transaction Matching — End-to-End Flow
Step 1 : Load Data Import transaction data from source and sub-systems into ARCS/NSAR
Step 2 : Auto Match Rule-based matching job runs; majority of transactions are matched automatically
Step 3 : Predict Match ML model evaluates remaining unmatched items and suggests predicted matches
Step 4 : Review & Confirm Users review confidence scores and confirm, adjust, or discard suggestions
Once the standard Auto Match job completes, a Predict Match job automatically runs to evaluate any remaining unmatched transactions. Users access predictions through a clean pop-up interface where they can review, confirm, or discard suggestions using simple navigation controls. The system even supports matches that require adjustments, pre-filling adjustment fields with your default rule settings.
Prediction Confidence Score
The AI assigns a confidence score to each predicted match, helping reviewers prioritize their attention. Reviewers can quickly prioritize high-confidence suggestions for fast-tracking, while giving more scrutiny to lower-confidence items
Key Benefits
Transaction Matching Assistance delivers value across three core dimensions — adaptability, efficiency, and control.
Continuous Learning & Adaptability
The ML model learns directly from your historical data and adapts to changing patterns over time as you periodically retrain it. Unlike static rules, it evolves with your business.
Efficiency & Time Savings
By predicting matches for exceptions outside your standard rules, the system drastically reduces the manual effort required from your accounting team during the close cycle.
Full Transparency & Control
Training and prediction tasks run as scheduled or manual jobs with detailed log files — giving administrators clear visibility into model performance and the ability to decide when to retrain. Users always retain full authority to confirm the best match or discard unsuitable AI suggestions. The AI assists; humans decide.
Prerequisites for Implementation
Before leveraging machine learning to predict your matches, three foundational setup steps are required.
-
1 Enable Predictive AI You must first activate the "Predictive AI" configuration setting within your Oracle ARCS or NSAR environment. This unlocks the machine learning capability at the system level.
-
2 Approved Match Types Prediction models can only be selected and run against fully approved match types. Ensure your match type configuration is complete and approved before attempting to train or run prediction jobs.
-
3 Historical Data & Initial Training Since the model learns from past behavior, you must have historical manual match data available to successfully run the initial "Train Match Prediction" job. The more historical data available, the better the initial model quality.
Best Practice
Monitor the volume of new matches continuously and periodically retrain the model to ensure predictions remain accurate and aligned with your current data trends. Business processes change — your model should too.
Ready to Elevate Your Reconciliation Process?
Contact Newarc Consulting today for expert EPM Account Reconciliation and NSAR implementations, support, and training.