Abstract

For big organization such as Goldman Sachs, valuable data that has large impact on the firm businesses is usually fragmented between multiple departments. The ideal data structure that the large organization should have is one single data base with customized data cut for each department. Large organization have been paying a lot of money and effort to reach this state, from large retail store trying to consolidate regional sale data base to online retailer like Amazon to enhance their data bank for analytical purposes. This dissertation will analyze Goldman Sachs data structure. We will go in depth to give the strength and weakness of the current structure and how it will limit the bank ability to adapt the changing environment. Lastly, we will discuss some changes that can be made to the current structure to improve those disadvantages and what benefit those changes will help the bank in current financial environment.

Back Ground

Company Background: Goldman Sachs is an American multinational Investment Bank who generates main revenue from Broker Dealer, Investment Management, Advisory and Investment Banking activities. Founded in 1869, the firm is headquartered in New York and as of 2014, the firm has offices in all major financial centers around the world. As of 2013, Goldman Sachs had 32,000 employees worldwide. In 2007, amid of Financial Crisis, Goldman open the commercial bank franchise headquartered in Salt Lake City, UT for private lending purposes. In 2010, Goldman agreed to the financial aid package from the Fed and to compliance with the regulator, the firm move the Bank’s headquarter from Salt Lake City to New York and become a bank holding entity.
Problem Background:
After becoming a bank holding entity, Goldman Sachs has to follow the Fed risk guideline which covers Market Risk, Credit Risk and Operational Risk. To be in compliance with this requirement, each banks need to send the Fed a consolidated report on those risk and the Fed will hold annual exam on each bank data to evaluate the accuracy and effectiveness of each bank. Banking industry received a huge blaming wave for its credit risk management during financial crisis and recent operational risk cases which result in billions of loss to the bank in market loss and fine. These events will directly damage the bank reputation and it’s not the risk that Goldman wants to take. To be in compliance with this new rule, Goldman Sachs is facing some challenges since the data was fragmentally stored across different department which would leave lot of room for mistake or discrepancies data between departments. For example, trade data is input by Operation, pricing is input by trading desk, risk management metric is input by Risk department, reporting is managed by Controller. Therefore, it’s not surprised to see different data for the same source in different system. For this reason, it’s very hard for management to make decision based on current data layout.

Solutions
The high level solution that Goldman Sachs can take is to consolidate all the data into one database that feed into other department. So instead of each departments storing and processing its individual data, there would be only one group in charge of input the data into the system while the other group is to either audit the data or use those input data to calculate their own department metric and upload it back to the database. For example, when there is a deal that got approved, it will first be booked in the system by IBD desk who input trade structure, pricing, valuation model, committee decision… Once the trade was approval, IBD will pass on this data to Operation who will maintain the current exposure that GS has in its book, mark of the trade and some other information: interest rate, collateral value… Once all the data was input by Ops, the data will flow into other departments system. Accounting department will see all the info and audit the trade to make sure the data is correct. Once they are comfortable with the data, Accounting Department can start using this data for their financial reporting purpose. Other end users of Ops data are Market Risk and Credit Risk department who will use the trade input and combined it with market data to run the risk model. If Market Risk or Credit Risk sees the risk to be above the limit, it will notice IBD desk to narrow down the risk and the data cycle will repeat to Operation and then to all other departments. This cycle may take a bit longer for the data to flow inside certain department but the overwhelming advantage here is that all the departments need to coordinate with each other to come up with an agreement on the current data.

Benefits The first advantage of the new structure is the accuracy of the data (Guorong, 2011). Since each data in the system need to go through multiple departments audit and everyone agree on the number that is in the system. Each department only looks at the certain aspect of the trade. Therefore, with the segmented database, only the metrics which the department interested in will be reviewed intensively. If there is any wrong data, the number will most likely be overwritten in that department alone without flowing to other impacted departments. This practice creates multiple issues such as the overwritten data might not be correct or there are discrepancies between each department reports. With the new approach, since all department need to coordinate and agree on the number in the system, the manager can feel comfortable that the number he looks at is consistent across the firm and there is less likely any error with the data. The second advantage is it will make the firm more efficient. One of the biggest issues that Goldman Sachs is facing right now is the inconsistency within the data from different department. The firm has to spend a lot of time to reconcile the data which is clearly not efficient and should be prevented in the first place. The whole practice of reconciling and fixing the data is the same as redoing all the work that has been done before which doesn’t add any value to the firm. The third advantage is if the input data is good, most likely the model output will be accurate. Goldman Sachs, like all other bank, depends a lot on stress testing to see how the firm portfolio will perform under certain scenario. The models need inputs from all departments. Since the stress models are complicated, if the input data isn’t accurate, it will be very hard for the user to know if the result is accurate and the firm may either react on the wrong result or miss out some opportunity (Canabaro, 201). When Goldman Sachs became the bank holding company in 2010, it also has to run CCAR stress test which directly impacts the firm’s ability to raise capital. Goldman Sachs not only has to run its own test but also has to send the input for the regulator for them to check the result. If there are any discrepancies between Controller’s data and Goldman Sachs data, it will be disclosed to public and will have negative impact on the firm. Last but not least, with the centralized data, the firm will enhance the security within the firm. The first security is avoiding data leakage internally. Each bank has a firewall system, nicked name Chinese Wall, prevents the data to flow from private side into public side to uphold the integrity of the market. However, there are a lot of times where the information cross the wall by mistake which result in the restriction of the public side to do business and may result in fine if the event is serious matter. With the new database, all departments have access to all information it needed. Therefore, there is no need for the information to be sent across departments. This approach will clearly define the data restriction for each department. The secondary data enhancement is that it will be easier to enhance the data security against external threat. (Shapiro, 1976)

Business approach The first step that the business needs to take is to build out a technology team who understands the data and is capable of managing the data. The centralized data that is described above sounds simple but in reality, it will be very complicated. There are almost hundred teams that contribute to the data or the end user of this data base. And sometimes, the data flow can be very complicated which requires a specialized team to be able to maintain or to look at any question arise from end users. The benefit of a support team for 24/7 support and to help out with the data streamline is the main factor for the success of this project (Grimes, 2009). With the current advanced technology level, it won’t be hard for the firm to set up this 24/7 support system. The second step is the firm need to create a robust structure. All the data in the central data is very important and will affect the firms business. We don’t want to get into a scenario where there is some glitch in the system which causes a hold across the firm even for a few seconds where there is no data in or out. This can result in millions of losses. There must be a backup system in place in case something goes wrong and the system needs to be reset to. Also, the firm should set up parallel data for maintenance and for tool testing purpose. Due to the importance of this data set, the firm needs to carefully plan out all the steps. The last step is all the large change need to have the support of the end user. The firm must clearly states why this practice is crucial for the survival of the company. The firm need to educate the employee about the new system so there is no operational issue when the new system is in place. Also, end user department need to be involved since the beginning from the designing phase to testing and implementing phase. By having end user to be involved into this process, it will help to make the project process faster, having fewer bugs and gaining support across the firm.

Business Process Changes With the new system in place, it would require a new process across all departments within the firm. The first process change is to make sure all the segmented data practice needs to be stop. This change might prove to be difficult since the firm is used to the old segmented data structure. This would require a lot of attention from the manager level to help the employee through the transition period. Education and constant monitoring are some practice that can help putting the team into new node. Technology issue is another set-back that can cause the user to resist switching to the new system. This issue can be mitigated with the help from technology department and with involving the users early on during the process. Another one time process that the firm needs to go through is to back fill all the old data into the centralized data base to create a data trend. All the data can only be useful if the end user can tell a story from it and one of the useful parts of the story is to compare it with the past data. This practice might prove to be very challenging since the company has been in global business for a long time (firm is founded in 1869) which results in a huge amount of data need to be back filled. When the data is backfilling, each departments need to go through the final data set and sign-off on the data. Last but not least, the new system would require the firm to run the data reconciliation process daily which is different than the current process. Each departments need to come up with a new process to be able to handle this new task since the amount of data need to be handling is large and the timing is fast which will make it quite challenging. Each department should come up with some surveillance method to automatically pick up names with issue or have high impact and focus on those names.

Conclusion With the increasingly challenging requirement from the Fed after Goldman Sachs become a bank holding entity, it becomes clear that Goldman Sachs need to transform its current data management system into streamline data base across the firm. The new database structure will not only help Goldman to satisfy the Fed requirement but also will make the firm to be more efficient and can quickly analyzing the data with the current market situation. The ability to make quick decision with the new system is crucial for the firm to be able to adapt itself in the fast changing financial market.
The main disadvantage of the new streamline data base system is that it will be costly and time consuming to set up, slow down the data flow and some resistance from the current employee with the new changes. However, these disadvantages will be completely offset by the over-whelming advantages the new system will bring. The new database structure will enhance the firm security, enhance data accuracy and efficiency, and give manager confident in the number and ability to make split second decision. There are strong evident that the organization with the better data quality will perform better than other company in the same industry (Gregory, 2011). For these reasons, I strongly recommend that Goldman Sachs should take a second look at their current data structure and implementing the change.
In high level, the firm needs to set up a new technology team to head the project and pulling in the end user in the early stage to help speeding up the process and reduce the system bug. While the new data base is building built, the firm need to come up with the streamline data procedure to ensure the system will be robust. At the same time, each department need to backfill the data from the old system to the new system format. This step is important since trend data is very important for the firm analysis. Throughout the process, the manager need to keep reminding employee the important of the new data on how it will help the firm and to provide training/supporting as employee is shifting toward the new system. Once the new data base is in place, it is necessary for the manager to provide constant attention to ensure the quality of the data that get input in the system and to manage employee’s expectation. The new technology team should also start expanding their role to not only able to understand the data base structure to help out with the user’s question but also be able to understand the metric maintaining in the database.
Through this project, it helps me to have better understanding about the important of project manager. This project also gives me a chance for me to apply what I learn in a real life problem and to come up with a technology process to mitigate the issue. I have learned how important the data will have on a company performance and what is the best practice to structure and manage a complicated database.

Reference:
- Gregory, A., & Hunter, K. (2011). Data governance - Protecting and unleashing the value of your customer data assetsStage 2: Fixing the priorities: Increase revenues and profits by raising your data quality IQ. Journal Of Direct, Data & Digital Marketing Practice, 13(1), 40. doi:10.1057/dddmp.2011.16

- Guorong, X. (2011). Data Processing Model of Bank Credit Evaluation System. Journal Of Software (1796217X), 6(7), 1241-1247. doi:10.4304/jsw.6.7.1241-1247

- Canabarro, E. (2014). Stress test design. Journal Of Risk Management In Financial Institutions, 7(1), 52-61.

- Turn, R., Shapiro, N. Z., & Juncosa, M. L. (1976). Privacy and Security in Centralized vs. Decentralized Databank Systems. Policy Sciences, 7(1), 17-29.

- Grimes, B. (2009). Guidelines to achieve centralized management. Networkworld Asia, 5(1), 16-17.