TD Bank Group has been using data collaboration software from a startup called Cinchy to solve a frequent problem for all banks: the need to share data among different applications internally.
Data integration is hard. It can consume enormous amounts of a technology team’s time and resources.
The first time the $1.1 trillion-asset company used the software, it was part of a proof-of-concept test to look at tying organizational master data together in one of its business units. It was able to connect reference data that was scattered across different systems written in different languages.
“We found it to be a great way to quickly tie data together, understand the differences in systems, and then use it as an exchange of data,” said Tim Clark, chief information officer of business segment technology at Toronto-based TD Bank. Clark is in charge of retail technology systems, including digital banking and back-office technology.
In other cases, TD has used the software to gather referrals from one part of the business, say a branch, and move it to another, such as wealth management.
“If a customer were to walk into one part of our business, looking for a particular piece of advice, and we think it's best suited to the wealth department, we can move that customer dataset across to wealth to advise them on assets or the products they're looking for,” Clark said. “If two systems need to talk to each other, the software can move the data between the two systems without fully translating it, and by marrying what the outbound data looks like and what the inbound data is going to need to look like.”
TD Bank often uses the software to pilot and prototype data exchange to see where issues might occur.
“Think about it as a fast method for us to compare datasets and to understand how we could connect them and enrich them as needed,” Clark said. “We can do that test to make sure that what we're doing is right, versus doing an elaborate project and then realizing maybe there's something missing at the end. For us as technologists, finding out whether something's going to work quickly is the most important thing when we're designing and delivering in an agile mode.”
What it does
The software itself is abstract and hard to describe. Cinchy CEO and co-founder Dan DeMers calls it an “internet of data.” It’s in a relatively new category of software called “data fabric,” which the research and advisory firm Gartner declared one of the top 10 data and analytics technology trends last year.
The software has built connectors that let users map their existing data stores and connect to new ones.
“Once a set of data is connected to Cinchy’s data network, the next project, user or system that needs that data no longer needs to go back to the original system of record,” DeMers said. “It has essentially a golden copy of that via the network. Conceptually it's kind of like if you buy a new computer, all you need to do is connect your computer to your network and that network is connected to the network of networks called the internet.”
The software can analyze data in disparate systems and visualize how they might work together. It can also be used to let applications share data, even if those applications are written in different computer languages and formatted differently.
“In the world of data sharing, what data collaboration typically means is I'm sending you a copy of my data,” DeMers said. “So two applications inside of a company share data by copying it. It could be over an application programming interface, it could be through [extract, transform, load] software. There are hundreds of complicated technologies to facilitate the copying of data.”
In Cinchy’s software, he said, two applications truly collaborate on data by working directly on it, not each separately getting copies of that data, the way two people can edit a Google document at the same time. Cinchy also has access, privacy and security controls built in.
“Dan has a vision around democratizing data, which loosely translates in my language as, how to make it seamless and easy to connect without using some of the traditional tools that most companies have spent a lot of time on: data warehouses, enterprise transformation layers,” Clark said. “Dan has a belief system that we can connect data more easily and more richly if you actually start to figure out where the relevant attributes of each data element are and how you actually would tie them together.”
A frequent hurdle to merging two sets of data is the fact that they’re often formatted and tagged differently. They might be written in different software languages. One database might have first name and last name in separate fields, another database has one field for full name.
Cinchy sorts the data elements and lets users tag them together, so first name and last name might be tied together.
“It's a slightly different take on how to translate,” Clark said. “Traditional enterprise transport layers would translate one field into another field and then make them all unique. Here, Cinchy just relates the fields together.”
The trouble with data integration
Many large banks use Cinchy's software to reduce the time and cost of delivery for projects they were already planning. Some are for customer personalization, some to get a 360-degree view of the customer, some for regulatory compliance.
DeMers worked at Citigroup for 11 years building enterprise systems, and he has also worked at Royal Bank of Canada. At the time, the bank had more than 10,000 different applications. He saw the data integration challenges banks cope with up close.
“There's no bank that has less than thousands of different applications,” DeMers said.
“The problem with that is the cost of integration is easily half of your entire IT change budget. I've personally been on massive projects that were nothing but glorified data copying machines. At the time, I would have called it a regulatory compliance project.”
Another problem companies run into with such projects is control.
“As soon as you copy data, you've lost control over it because today the data doesn't help protect itself; it needs the application to protect it,” DeMers said. “So if you have a system that is managing customer data, maybe it's doing onboarding and know-your-customer compliance. If it is the system of origination, and now I copy a slice of it and put it in my transactional system, the logic that was protecting that in its system of origination is no longer applicable.”
In a bank, data might be copied and sent among 5,000 different systems.
“Some of that is highly sensitive, personally identifiable information,” DeMers said. “When you think of the growing awareness of data controls, that is a massive problem.”
