Pershing, which provides securities clearing, execution and settlement services to institutional investors and investment advisors, has greatly expanded its use of business intelligence in the past few years and is planning to further expand its focus on BI. Patrick Yip, a director at the company, talked to Computerworld US about the implementation.
You have scaled your BI tools over the past several years to 78,000 users. What have been the key factors in allowing you to accomplish that?
You have to find someone who has done this before — especially when you have a lot of users hitting your database with a lot of queries. You need to work with [lines of business] to convince them that you can’t have everything. You can’t sit in front of this reporting tool and say you want to query everything. You have to put some parameters around it so you can optimise it.
We treat this as a mission-critical application because if there is an outage, it automatically goes through a structured, managed process. It doesn’t take long before the CIO and sometimes the CEO get on the call.
With so many users accessing the system how do you monitor it for usage and for performance?
A lot of traditional BI applications have been treated as second-class citizens from the monitoring standpoint. We have our own tools we use to monitor end-to-end use [of Information Builders’] Resource Analyser on the back-end, so we know which databases and schema are being accessed. We have our in-house monitoring tool to make sure the loads to the data warehouse are completed on time. In real time, we do have a pretty formal process. Every morning to make sure the application is working and that the analytic reports are coming up the way they should be. We have weekly and monthly scorecards, and we’re always looking at usage. We break it down by the top brokers and dealer firms that are using the system.
How are you expanding the way you use the BI tools?
It is becoming more and more mainstream and embedded in business folks’ day-to-day operations. We have applications that are taking data at near real time and doing transformation on the fly — such as requesting a cheque or a federal funds wire. We have the BI applications that help managers identify throughout the day how many of those transactions are being requested, how many have been processed and how many are still pending or have been there for several hours.
What are some of the lessons learned around BI?
It is more important for BI applications to get out there sooner rather than later. A lot of projects traditionally have taken a long time to finalise the requirements and get out there. Get BI implementations out there as quickly as possible. The 80/20 rule works pretty well. The day you put it into production the requirements will change.
We have quarterly release for reports: make sure you measure what you are doing. We have made tremendous performance improvements on some of the reports ... to the tune of 500%.