itemeditorimage_5b9beaf50fa08

Enriched Data for Better Banking Decisions

Transaction data is the most insightful data that financial institutions can harness in order to better understand their customers. The rich data from transactions, particularly from mobile devices, is fueling the growth of a new breed of personal financial management. However, banks and their customers have some distance to cover with respect to harnessing this data effectively. According to McKinsey & Company, the amount of meaningful data has increased significantly. By 2020, about 1.7 megabytes of information will be created every second for every human being on the planet[1]. Imagine the available amount of valuable data, and its potential to transform the consumer experience and businesses. Among all industries, the banking industry is considered the most data-driven of all. Regulatory and insurance requirements require banks to store many years of transaction data. More recently, digital technologies are at the helm of transforming legacy banking operations - but can we enrich data to create business value? Transaction data has immense potential to improve a bank’s relationship with its customer. Banks will be able to comprehend billions of transactions across locations to understand customer behavior and buying patterns based on parameters such as gender, location, and time. The same data could help banks’ merchant customers improve their businesses by analyzing the transactions of their consumers. Lastly, the information is useful to key entities in the industry. If used wisely, based on transactions alone, data can help analysts assess growth of banks and/or their institutional customers. Such is the power of enriched data. Recently, for a top global bank, Envestnet | Yodlee harnessed data to better understand customer behavior, payment preferences, share of wallet, and even what financial services are best suited for specific customer segments. We understand that enriched data is the most critical. Transaction Data Enrichment (TDE) is a machine-learning engine that provides data based on its findings from millions of transactions. TDE allows for the delivery of contextual information to customers, while protecting their data. Its accuracy and depth leads to an evolved user experience and ability to make better financial recommendations. In the context of transaction data, banks usually need three categories of data to generate value: merchant name, date and amount of transaction. Artificial Intelligence and Machine Learning support speedy and accurate analysis based on the categories and its self-learning capabilities allow constant analysis of billions of bank transactions. Consumers find it difficult to track transactions based on merchant identities. A more simplistic explanation to the issue – Envestnet |Yodlee captures variations in merchant names (e.g. “PizzaDom” or “Domino” or “Pizza”) and also acts as classifier to determine if there is any variation of category-indicative key phrases present in the transaction which may overrule merchant based transactions. This allows a significant elevation in user-experience. For example, an app based system can instantly display, to the customer, the various transactions and expenses within a certain category. They will understand how much they spent on pizza  and if they must cut down their monthly expense on eating out. For the bank itself, the enriched and contextualized data makes it easy for sales & marketing and other internal teams to:

  • Identify cross-sell and upsell opportunities
  • Target the right customers with relevant and timely campaigns and promotions
  • Deliver personalized financial advice and engage with customers in a more meaningful way

The key to Envestnet | Yodlee solutions is a robust analytics led offering. A good structured learning and prediction module is based on a sequence to sequence model architecture that demonstrates success in a variety of tasks such as machine translation, speech recognition, and text summarization. The next step predicts the category of given transaction in two parts: (a) Merchant-driven, and (b) merchant-agnostic. This kind of structural modelling gives us the flexibility and scale to predict as many as a couple of million merchants. Coupled with a specialized hashing module for merchant prediction, it ensures that the model is highly accurate with top merchant names that occur more often. This also provides a systemic flexibility to identify local merchants. Вход на сайт Вавада официальные рабочие зеркалаFinally, Envestnet | Yodlee’s big data engineering stack ensures that the Machine Learning outputs merge seamlessly with web interfaces to provide a quality user experience. One of the major strengths of deep learning models is its automated process that also learns while they perform. It would otherwise require huge human effort to go through and write the transactions manually. Envestnet | Yodlee has invested a sustainable web service-based solution which can handle tens of millions of transactions daily in a reliable and time effective manner. Over the last few years, Envestnet | Yodlee’s TDE has achieved an accuracy of over 90%, while having identified over two million unique merchants by populating more than 90% of the transactions. We are able to accurately identify the top 2000 most frequently occurring merchants with over 99% accuracy. TDE allows banks to turn ambiguous transaction information into clear, contextualized data. While it allows financial institutions to extract deep insights into customer needs and behavior, it will always be the bank’s intent to put insights into action that will elevate them into the next phase of growth.   [1] https://www.mckinsey.com/industries/financial-services/our-insights/analytics-in-banking-time-to-realize-the-value. Read more about our Data Enrichment products.