Manual matching, a hidden problem?
Reconciliations are an integral process in any organization. They ensure the accuracy and integrity of financial data by comparing and records from different sources and matching those which originate from a common business event.
The traditional industry approach for matching these transactions is to use an ordered set of rules to match transactions together based on their attributes.
For many reconciliations where there are many fields which can be used for matching and where unique references are available or data quality is high, traditional rules-based matching engines can deliver excellent match rates.
However, for other kinds of reconciliation (e.g. cash recs, GL, settlement recs), a lack of unique references and or limited number of data fields with imperfect data quality, rules-based matching can hit a limit, leaving recons teams having to manual match thousands of transactions daily.
So, although the average match rate across all of your different reconciliations may be high, those particular reconciliations which have a lower match rate may be absorbing a lot of your team's time and producing process bottlenecks. (According to the Pareto principle, 20% of causes determine 80% of problems!)
Hence just monitoring overall match rates may keep this issue hidden (as well as the fact that in many organizations the manual matching is performed by offshore teams who may not have direct communications with the high level operations management team).
The prevalence of manual matching work is borne out by our surveys across industry members which show that manual matching takes an average of 30% of the recon team's day. This can lead to team burnout, increased operational costs and delayed financial reporting, causing a lack of real-time visibility into financial health.
It is instructive to estimate how much time your team is spending on manual match activities for these recs (to help estimate this for your organization, see our simple survey designed to discover the major activities being performed (https://www.surveymonkey.com/r/3GZD9WN). Operartis’ industry surveys using this assessment performed across 40 financial institutions showed that reconciliation teams are spending on average 30% of their day on manual matching because of these problematic reconciliations.Â
How does machine learning help?
Those reconciliations requiring significant manual matching are the ones where machine learning can be usefully deployed to boost the automated match rates and reduce the user workload. The ability of machine learning to learn data patterns from historical matches allows the machine learning system to create matches based on those patterns without the need to rely on the kinds of rigid matching logic which rules require.
The problem of matching pairs of transactions is a form of classification problem where the class of each given sample must be predicted. In the case of transaction matching the sample consists of a transaction of one type (for example a bank transaction) and the class consists of the ids of one or more transactions of the opposite type (for example a GL entry) with which this transaction should be matched. (or if not ids, at least the key characteristics of those transactions should be predicted).
Machine learning can be supervised (correct classification examples are available to learn from), or unsupervised (correct classified examples are not available). For reconciliations already in production, historical matches are available. Once the full reconciliation operational process has been completed the majority of these matches can be considered as correct classification examples. This allows the user of supervised machine learning algorithms to learn the data patterns which exist between transactions belonging to each historical match. Once these data patterns are learned, the machine learning classifier uses these learned patterns to predict the matching transactions for a given (sample) transaction.
Such machine learning can be used as a standalone matching engine, or as a booster to supplement rules-based matching to provide improved combined match rates.
The benefits of incorporating machine learning into your organization's reconciliations are multifold:
Increased productivity:Â Reducing the man-hours required for manual matching. This allows your finance team to focus on higher-value activities like financial analysis and strategic planning.
Reduced process time: reducing the manual matching workload means that reconciliation teams can complete their reconciliations faster, ensuring more timely financial reporting.
Reduced mismatch risk during the manual reconciliation process. Users can make match errors, especially if they have to rush to complete their manual matching activities. This can lead to potential financial or audit issues.
Auto-configuration :Â ML based systems build their matching logic automatically from historical match data, auto-calibrating to ensure correctness. This removes the need to set up large numbers of handcrafted rules and regularly maintain them.
Auto-maintenance: Effective ML based systems include automated performance monitoring and auto-retraining to ensure that matching continues to provide both optimal match rates and optimal accuracy.
Unlock the Power of Reconciliation Automation
Are you ready to transform your reconciliation processes? Matchimus offers a compelling solution that delivers unmatched efficiency, accuracy, and intelligence.
Get a Quote Personalized to Your Use Case with Our Proof of Value (PoV) Assessment.Â
We'll measure the match rate improvement on your reconciliation data and provide an automation report detailing your ROI before you commit to purchase.
Let's talk it over – schedule your demo to increase efficiency, improve exception management, reduce costs, and enhance visibility into your financial data. Matchimus – the future of reconciliation is here.