Analytics Democratization vs. Governance: Are They at Odds?
Democratizing Analytics at UBS
Well, are they?
Are you finding that governance and democratization are at odds in your organization? Are your line-of-business managers slow to adopt analytics in general and Alteryx in particular because they see more risk than reward? How do you overcome that resistance while still maintaining compliance with regulations in your industry?
I had a fruitful conversation with Nick Bignell, who runs the data science service at UBS, a bank headquartered in Zurich. UBS has 100,000 employees and outsourced individuals across offices in more than 50 countries. As the largest wealth manager on the planet, UBS specializes in investment banking, asset management and financial services. The company was founded in 1860 and has a lot of data — some of it going back that far. Here are Nick’s useful insights for global companies, that also apply to the analytics efforts in companies of all sizes.
Democratizing analytics at UBS
The first thing Nick wants you to know about the world of finance is that it depends (too) heavily on spreadsheets.
“All banks run on Excel,” he says, “and if anyone tells you anything different, they’re fibbing.”
His point is that there’s too much low-level, manual data prep work going on in financial firms and, really, in organizations everywhere. That’s why self-service data preparation was one of the two basic capabilities he wanted to boost at UBS. He and his team wanted to lift that work out of the 1990s and bring it into the 2020s. They decided to focus on upskilling their fellow employees in data literacy so that they could add value to the company and to themselves.
The other capability his team wanted to expand was self-service machine learning. That has involved bridging the gap between the data scientists at one end and the business-area knowledge at the other end. They wanted to work with a tool that would enable knowledge workers to apply statistical analysis and get the most out of their data. The goal was to make better decisions more quickly.
Within UBS, those two capabilities combined are called the Data Science Service, aimed at creating a platform for making data and analytics available to the entire organization.
Smoothing democratization into the company To really democratize analytics, Nick recommends:
- Train-the-trainer training
- A center of excellence (COE)
- A community
- Low-friction licensing for easy onboarding
When you think about talent, data scientists are rare, and data scientists who also know how to use your chosen analytics platform are rarer still. So, UBS decided to take system engineers and train them in data science, then let them build from there. In effect, they created that skill set and capability in-house.
Nick’s team also focused on training, which was no small task in a company with 5,000 Alteryx users. They have developed a train-the-trainer model that relies heavily on partners and vendors who can train UBS users.
But if you want to have a successful analytics tool and capability, you’ve got to invest in a team of platform specialists who are going to support it. Don’t tell yourself that it will suffice to post the tool on a portal someplace, and that users will spontaneously adopt it. UBS provided a center of excellence around Alteryx and similar tools to attract users who were developing their skills. Those budding experts used their knowledge to enable, train and help other users.
Similarly, they created an internal community in which users can talk to and learn from one another, and where Nick’s team can announce and publish new work. Over the past few years, they have cultivated almost a dozen champions in different departments around the organization. The champions promote Alteryx and other tools within their particular department and introduce it to the business. Finally, UBS has lowered the barrier to adoption by packaging Alteryx so that license activation is automatic. Using the capabilities in Alteryx’s local licensing server, UBS has enabled users to simply go to an internal software portal and request access to the software. Following approvals from a user’s line manager and a member of Nick’s team, which can take as little as a few minutes, users can start Alteryx at their next login.
The tool takes root
At UBS, Alteryx started getting traction within small groups, first in Operations, which had a few use cases ideally suited to the tool. But the big aha-moment came in Finance, where Alteryx became known as “the reconciliation tool.”
Banks need to reconcile the transactions that go through an exchange or trading platform against transactions cleared through the settlement engines. In case of errors, someone from Operations fixes them. There’s been a long legacy of expensive applications to handle reconciliation. But most people who use Alteryx know that it can easily handle reconciliation, and a great deal more. UBS users discovered they could apply a depreciation module to the reconciliation use case, and the potential of Alteryx shone through. “Users suddenly understood, ‘I can use it for that. I can use it for this,’” says Nick. “And it just grew as Finance and Operations built programs around pushing Alteryx out to their groups.”
Analytics Democratization and Governance
In a heavily regulated industry like banking, you’re always waiting for the other shoe to drop. So, once Nick had made Alteryx and related tools available to thousands of UBS users, he knew he’d get questions from the information security officers. “How is this controlled? Can people just connect to anything and do what they want with the data and push it out?” Of course, there are controls, says Nick. His team has narrowed it down to three scenarios where the users of the tool have to abide by the rules. And, fortunately, Alteryx makes it easy to track user activity.
1. Workflows running on the desktop
Most of the time, users sit at their workstation, run Alteryx Designer on their desktop and create a workflow that executes regularly. How does UBS govern that?
For one thing, UBS doesn’t allow Scheduling on the desktop, because users work on a virtual desktop. The virtual machine is scrapped every weekend and replaced the following week. But Nick’s team can identify the activity and source by the phone-home metadata going from the application back to Alteryx. So, they capture that and save it to an internal MongoDB database. That’s important for governance because they can tell who’s running what even though Alteryx is running as a desktop application on their workstation. The result is an inventory database of all activity, from basic usage like data prep up to using Python code or the predictive analytics suite of tools. Nick’s team can point them to the end user application policies and get them to affirm that they will follow company guidelines.
2. Shared, ready-made applications
Some users at UBS create applications in Alteryx and place them on a server for other users to use. These “artisans” have created and uploaded about 2,000 workflows to a server. A population of about 8,000 other users can access the workflows, but they don’t run Alteryx Designer; they use Alteryx only to interact with the ready-made applications on the server. The people who create those applications have created code around their solutions. Even if they’ve simply dragged and dropped, they’ve created something used by other people in the organization for a business purpose, and it’s easy to track that.
3. Automating and governing workflows with server
Because the local desktop is virtual and does not permit Scheduling, users create and test some workflows locally, then upload them to the server to run regularly. Activity on the server is very easy to track.
For all three of those scenarios, UBS maintains separate inventories. Users are instructed to follow all applicable rules and controls.
Providing support for user-created workflows and applications
Within the business streams that provide the applications, specific areas within UBS offer their own functional support. The automation support team for each area also supports the Alteryx workflow with Level One support for applications created on the server. Next, Nick’s team is split between Level Two production support and Level Three engineering support. Level Four is escalation to Alteryx, for problems UBS has never come across before.
Spreading costs fairly
“To be frank,” says Nick, “one of the main problems is figuring out how to pay for Alteryx and other analytics tools.”
It’s straightforward in companies that keep the tool in a silo, like Finance or Marketing. But UBS wanted to use Alteryx companywide. They also wanted to avoid having to continually order new licenses for Alteryx every time somebody else in the company wanted to try it. That meant working out a way to charge back to different parts of the business depending on usage.
Nick says that UBS’s secret to being able to grow with Alteryx is to have more licenses than they have actual users. “That way, we’re not ordering new licenses every five minutes.”
They fund the extra licenses by charging back an extra 10% to the existing license users. They take last year’s usage, which is recorded in the MongoDBs on their licensing server, and compute the charge on an annual basis. Then, when they receive the invoice from Alteryx, they adjust the ledger to push that cost out to the users in the business.
Everybody who uses Alteryx activates a license, which is plainly visible, so Nick’s team charges them. “If they say, ‘What is this thing and why am I being charged?’ we tell them, ‘Ten people in your area used it. Here are the logs that show the license activations.’” The model allows for trials. Nick’s team applies the charge only to those who used the tool for more than 120 days in the previous year. So, if a group decides Alteryx is not for them, or they didn’t get time to test it, or they’ve changed priorities, they can uninstall Alteryx after a few months. In that case, there’s no charge in the following year.
Conclusion: The blurring between IT and the business
Nick thinks the time is right to eliminate the distinction between IT and the business.
“There’s a blurring happening,” he says, “especially as self-service tools like Alteryx become more widespread. The tools are attractive to anybody on the cloud journey because they make the lines between IT and the business fuzzy, the way they should be.”
Find out more from our success story, “UBS AG: On Governance of Models and Workflows in a Regulated Environment.”
This article originally appeared on #INPUT by Alteryx
https://www.alteryx.com/input/blog/democratizationvsgovernance
