A Billion-Dollar Problem Faced By Financial Institutions

Financial institutions’ trade data management systems, whether related to banking or trading, are undergoing changes with an unprecedented expansion in market information and corresponding compliance pressure.

The challenges faced by financial institutions to manage and reconcile trade data include exponential volume growth in traded products, suboptimal reconciliation infrastructure, IT application silos that dot the landscape of securities firms, and the growing complexity of products

On top of all these technical difficulties, government regulations will only get more stringent and broader, with strong penalties for non-compliance. This drives the need for effective solutions for trade reconciliation to have smooth ongoing operations and robust risk analytics. Many firms fill in gaps by hiring more people focused on compliance and operations, but it’s not a scalable solution.

Machine Learning for Risk Analytics and Trade Reconciliation

Tamr’s machine learning-based platform, Unify, enables firms to solve transaction data challenges without overhauling their existing infrastructure and can address issues such as: low auto-match rates and heightened risk from unmatched items, long turnaround time and communications gaps, decreased trading agility because of the usage of dated data, errors due to multiple hand-off points and systems, valuable staff resources that could be spent on risk control. In short, Tamr Unify improves your data quality and analytics to achieve better decision making and greater capital efficiency.

Legacy solutions can’t scale to meet your needs

Trade reconciliation is no easy task. Obtaining a standardized set of reconciliation data from all systems just doesn’t happen. Firms often need to integrate their internal account system, treasury and securities, their portfolio system and cash management, etc, creating a portfolio of reconciliation systems and maybe even a “reconcile-the-reconciliation” scenario.

Even after the data is integrated from different systems, legacy solutions approach trade data reconciliations by utilizing user-configured rules-based matching logic (including Excel spreadsheets).

But rules will need to change over time, the audit trail needs to “remember” the rules, and the rules need to be able to handle tolerances. As a result, the rules create a heavy maintenance burden. With the ever-increasing volume of transaction data, these manually created rules and scripts simply can’t keep up and are not reliable.

Human-Guided Machine Learning Approach

Tamr takes a novel, human-guided machine learning approach to integrating and matching transaction data. In both the integration and matching process, Tamr Unify’s algorithms learn from internal experts that best know the data.

By asking a series of “yes or no” questions on what constitutes a schema match or record match, Tamr will utilize this insight to employ algorithms and take a pass at the datasets with the goal of identifying potential matches of attributes and/or records. This approach is far more scalable than a manual or semi-manual system to deliver the operational control and efficiency.

Tamr Unify is able to identify the two records to be matching even though there are obvious misspellings, truncations, missing fields and different securities descriptions – a task that a rule-based system won’t be able to achieve easily – if at all.

More importantly, Tamr achieved this level of auto-matching all through “learning” from internal experts, without data engineers having to write any rule scripts. Overtime, with the occasional feedbacks provided by data users, the system will get better at recognizing transactions.

Advantages of Supervised Learning Approach

Tamr’s machine learning-based approach to matching transaction data at scale has numerous advantages over traditional rules-based approaches, and is perfectly suited for large financial institutions. Reports built on top of Tamr’s unified datasets are trustworthy due to the inherent engagement with business experts within the technological processes.

Tamr interoperates with existing systems and can be used to model data from any domain. This is in stark contrast to rules-based approaches that only work with certain IT systems. It can save effort in 90% of the processes associated with manual reconciliation, resulting data arrives in a matter of days instead of months, enabling timely decision making.

Tamr delivers significant value at a fraction of the upfront cost of traditional approaches. Moreover, traditional rules-based approaches often increase exponentially in cost as datasets are added to projects. And automated workflow derives process convergence, resolution and reduction of errors, increases productivity and compliance adherence.

Conclusion

By utilizing machine learning for transaction matching, financial institutions can realize material benefits in the form of increased productivity, greater scalability, streamlined reconciliation processes, enhanced compliance and improved client service.

With today’s heightened regulatory scrutiny, reconciliation and exception matching is no longer just a business issue but has a material compliance angle to it. While traditional data approaches fail, machine learning capabilities can be leveraged to rapidly meet business and regulatory requirements.