+1-212-810-0264
+1-212-810-0264
BenchRec
Driving reconciliations operational excellence through benchmarks
BenchRec has been launched as a collaborative industry effort to provide datasets and operational metrics to allow for the development and comparison of both rules and machine learning based reconciliation approaches.
Organizations looking to optimize their reconciliations processes can use the datasets to drive insights into their operational reconciliations through the measurements of match rate, match precision and match confidence calibration.
Academic members, solution providers and financial services firms are invited to participate in the benchmark and where possible to donate additional benchmark datasets.
​
The first ever reconciliation benchmark was launched by Operartis at the 2023 ACM AI in Finance conference (https://ai-finance.org/icaif-23-competitions-datasets/) and now has been made available here to all interested parties as the first dataset under the BenchRec initiative.
​
The training and evaluation datasets contain obfuscated sets of matched GL and bank transactional data taken from a production reconciliation from a Tier 1 financial institution. The dataset has undergone validation from multiple teams to ensure it is sufficient to build an effective rules or machine learning model matching engine.
​
From academic labs to financial services firms to reconciliations solution providers, everyone is welcome to sign up and participate. The forthcoming public-facing leaderboard is designed to encourage innovation and collaborative understanding of this ubiquitous industry use-case.