Tamr’s cloud-native customer master data management (Customer MDM) seeks to create a unified, accurate, and persistent set of unique identifiers and attributes that wholistically depict a customer in order to connect millions of valuable customer records across multiple organizational silos, customer systems (like Salesforce, SAP, Hubspot, etc.), business processes and units.
While customer MDM is one of the most effective mastering projects if you’re needing to deliver value in short order — banks and other financial institutions have also identified other critical entities that need mastering. These mastering initiatives rely on Tamr’s data mastering capabilities in a similar way to customer mastering, and also achieve business outcomes.
Firm A is a California-based financial institution licensed in all 50 states and operates direct to consumer, in-market, and wholesale business channels servicing customers across the nation. Their main business line is to service home purchases and refinance loans nationwide. The company funds more than $300 billion and ranks as one of the three largest retail non-bank lenders and one of the leading retail mortgage lenders in the United States.
Our use case champion at Firm A identified the best entity, Real Estate Agents, as the mastering initiative with the most immediate business value. Our sponsor in the company already identified the 80/20 rule for the Agents. They knew they could gain the greatest value by identifying the best-performing agents and marketing to them.
Since Tamr has a better solution than the IT team could create themselves, Firm A decided to try using Tamr to master their Agents data. A dataset of roughly 660,000 Agents in California was used for the technical proof, and Tamr was able to provide mastered results in two days. With confidence in the earlier technical proof, Tamr is now being deployed on Firm A’s instance of Azure to master for both buying and selling agents with roughly 1.6 Million records.
Bank B is a consumer financial services company with annual revenues over $15 billion and a market capitalization of $30 billion. The bank offers consumer financing products, including credit, promotional financing and loyalty programs, installment lending to industries, and FDIC-insured consumer savings products. As one of the largest providers of credit cards in the U.S., Bank B expanded its auto and home acceptance locations and value propositions, and they can be used by consumers to finance their home needs at more than a million retail locations nationwide. They also increased its Car Care acceptance network to more than 500,000 locations across 25 categories, including gas, auto parts and service, car washes, parking, and ride-sharing.
The data problem arises when multiple sources of merchant data must be matched for downstream purposes, like operational use cases for customer services and analytical use cases for marketing efforts, etc. These critical but hard-to-work data sets have varying formats, naming conventions, and levels of data quality, making it difficult to match within and across sources. The matching process relies on a third-party data vendor’s matching algorithm, and anything not matched by that process needs manual review. Bank B needed to dramatically improve its process to be faster, more accurate, and require less manual effort.
Using Tamr’s machine-learning, human-guided approach to mastering, this company matched merchant records within and across systems. These matches will be at the location level. The data will be used downstream to serve its customers better, such as allowing cardholders to find locations to receive preferred financing. Tamr was selected because we have demonstrated the power of machine learning to quickly and accurately master home and auto merchants, as well as enrich records with third-party datasets.
The Blackstone Group Inc. is an American alternative investment management company based in New York City. Blackstone’s private equity business has been one of the largest investors in leveraged buyouts in the last three decades, while its real estate business has actively acquired commercial real estate, making them the world’s biggest corporate landlord. As of 2020, the company’s total assets under management were approximately US $619 billion.
Blackstone has successfully implemented several mastering models around their Portfolio companies, which provide enriched, mastered data to their customers. Blackstone fully stood up and implemented Tamr with a team of two people and reached production on AWS native in 50 days, enabling downstream of data consumption in Snowflake.
Data is the core to Blackstone’s investment strategies and due diligence, and Tamr is a lynchpin to their data strategy. As one of the largest real estate developers and investors, Blackstone has their MDM system for the properties they are evaluating or managing. By applying the same algorithm of customer mastering, Blackstone can manage and enrich property data at the site level. This is the fifth use case Blackstone has implemented with Tamr. Learn more about how Blackstone is using Tamr here.
Why Machine Learning Versus a Rules-Based Approach is so Versatile
Today at banks and other financial institutions, data is growing exponentially, and variety is constantly changing. With rules-based data mastering systems, the rules need to eliminate exceptions, express unique identity and pass the concept to each system. The number of rules needed to manage ever-growing sets of data becomes unmanageable once the number of data sources reaches around 20 systems, according to Tamr Co-founder and Turing award winner, Dr. Stonebreaker. Additionally, missing data, the temporality of data, and schema changes can easily break the system. And this is just from the technology perspective. From a people and process perspective, rule-based systems have a hard time capturing team knowledge and institutionalizing it, resulting in new data engineers unable to interpret lengthy coding from before, or the data engineers maintaining the rules can’t understand the business context.
The human-guided machine learning approach, however, is driven by a statistical model that takes into consideration the entire dataset. The data models are refined as more data is supplied, creating a system for constant improvements: the bigger the data sets, the more accurate the results. Machine learning leverages human subject matter expertise to learn over time, thus capturing their knowledge and building it into the algorithm without any coding. The models produce a confidence level that determines if a match has occurred, regardless of the entity mastered, and output can be manually adjusted, as data is complicated and can be inconsistent at times. This is why customers can utilize Tamr in such a versatile manner and employ Tamr’s Customer Mastering solution creatively to achieve their business goals.
Their mastered data is continuously ‘live’ and up-to-date and can be fed to operational and analytical systems to drive business outcomes. With a human-guided machine learning solution such as Tamr’s, ROI can be dramatically improved (643% three-year RoI’s, as outlined in a Forrester Total Economic Impact Study).