“Our development teams move at light speed here,” Mike Burkes, VP of Business Intelligence, Fraud Management, and Revenue Assurance at Ultra Mobile says, “Decisions made on a Friday are worked into the next available Scrum sprint and can be out the door to our customers in as little as two weeks.”
Ultra Mobile provides international voice, text, data, and cell phone service that caters to value-seeking foreign nationals living in the United States. The company started in late 2012 and by 2015 made the Inc 500 ranking as the fastest growing private company in America. “Basically, our products are our programmed onto a SIM card that provides international calling, and that SIM card comes at a very low cost.” Burkes joined Ultra Mobile in 2016 after 16 years running the data department for Toyota’s North American Sales headquarters.
Ultra Mobile is a mobile virtual network operator, which means that they run under a major US carrier’s network. Since inception, Ultra Mobile’s focus has been on providing international cell phone service to foreign nationals living in the United States – a potential market of about 40 million people. In late 2016 Ultra Mobile launched the MintSIM brand which introduced the concept of bulk pricing for pre-paid wireless with an all-digital footprint, delivering low pricing for the domestic market.
The Need for a Business Intelligence (BI) Upgrade
Ultra Mobile’s massive growth, coupled with the importance of responding in a quickly changing market, created massive amounts of data from disparate sources, yet a rigid Data
Architecture hampered business users from easily accessing relevant insights from that data. The business was evolving faster than their Business Intelligence capabilities, he said:
“It’s basically a PHP back end, and coding had gone to a lot of silos of data. When I joined we had made some steps to move into a Data Warehouse, but everything was operationally being done in SQL. There were SQL reporting tables on a different server that were moved into the Data Warehouse for deep dive analytics, but largely they had done everything in Microsoft SQL Reporting Services, where all of the data was eventually moved back into SQL for reporting. Ultra sent out about 140 reports a day via e-mail to their different business entities.”
Within six months of Burkes’ arrival, Ultra Mobile had also changed their CRM system, moved to a ZenDesk ticketing system, and migrated major applications from SQL server to Aurora. “If you look at the background of changes that are happening and continue happening as we grow, the data on the back end is just – it’s chaos.”
With “no good way to effectively manage it or govern it,” Burkes saw his first priority as modernizing their BI tool so business users could get meaningful reports. “The main complaint I got when I came in and surveyed everyone was that they just didn’t want to read the reports anymore.” With reports arriving every hour, and with no relevant context, “People were ignoring the information that was coming in. They were inundated.”
Burkes considered multiple options to modernize the front end but most were built on SQL constructs, which required aggregating data to pull from a single source. Instead, he wanted to modernize the Business Intelligence Architecture and find a more sustainable way to maintain a single source of truth.
Data Virtualization as a Solution
Burkes had worked with Denodo at Toyota, “When we were having trouble getting the security to work out of the systems in Hadoop.” For Ultra Mobile, he wanted a business abstraction layer that would create a single source of truth without the cumbersome process of aggregation. “I brought it in as a way to kind of harmonize all the data behind the back end so they would have a consistent front for the front-end tools.”
Ravi Shankar, CMO of Denodo, says that Denodo has been providing Data Virtualization technology for 17 years. Data Virtualization allows end users to access data from disparate sources, locations, and formats without replicating that data and without being concerned with where the data originated. According to Shankar,
“The consumers just go to the Data Virtualization layer and ask for the information. Data Virtualization goes and figures out that information and brings it back from the different sources, which are all in different formats.”
Burkes now uses Denodo to cache historical events and marry them with real-time data to provide more meaningful insights. Caching the last 29 days and 23 hours of detail call data, Ultra Mobile streams in the last hour real-time, as the Denodo layer allows them to combine real-time data with the cache and present it as one 24-hour block in Looker. “That was something that Denodo helped us work through – to figure out how to run the SQL so that we could traverse the persistent and cached data transparently.”
It’s now possible “To come in and quickly look across every country that we service and see the call quality analytics in real-time.” Burkes said:
“When you peel it back a layer, the special sauce of our business is the optimization of international calling cost by leveraging local ‘in-country’ markets to drive the best cost. This competition for services drives reduced pricing and allows us to build margins through preferred priced routing. Low cost and quality aren’t always partners so we’re always conscious of the switching is doing to quality on the front end of a call – we have to carefully tune that engine and make sure we’re not overriding some small provider that may be giving us a great rate but causing performance issues.”
Impact of the Solution
Ultra Mobile has been able to track KPIs and other drivers, and can now launch new initiatives with less risk because their reporting is current and relevant, he said. “It’s allowed us to improve our time to market and increase the quality of what we’re doing.” They can launch on a Tuesday, “and on Wednesday we’ll have full-fledged reporting with history,” providing benefits to end users from the top down. “The team manages all of the chaos in Denodo, so it’s just very quick for us to manage changes,” and keep that chaos from reaching the front end. This is relevant, Burkes said, because, “Our biggest user is our company president. He uses it daily and is a great champion of data-driven decision making.”
Burkes remembers when it took three days or more to get deep analytics out of a reporting tool.
“It’s incredible to be able to take a several billion-row data warehousing table and tie that with a relatively small SQL table, and see seamless performance between the two because virtualization is actually working, and because of the optimization of how the queries are run, and [because] you’re not moving data.”
Improving the reporting process has also created a change in attitude toward reports. “It’s moved us from a push of reports to now, [where] people go in and pull reports.” The team tracks progress on removing the reports via a leader board in Burke’s office. Over 120 of the original reports have been retired or re-invented in Looker.
“Being a small company, there’s a lot of hard number crunching that’s required to sustain growth and profitability – we’ve really got our precision and accuracy down to a science here. We forecast our invoices based on projections of usage and that helps to prevent surprises and expedites our reconciliation. All possible because we’ve got our data in reasonable shape now, from the business layer forward.”
Besides increased efficiencies and better end user engagement with reports, Burkes also says he believes the parent company has recognized improvements in the accuracy of Ultra Mobile’s data. “I will say that they’re trusting the information that we’re getting, and I think we are faster at making changes.” Ultra Mobile’s provider has been undergoing a nationwide revision of their systems, he said, and now, “We’re kind of leading. We’re waiting on them to make changes instead of the other way around,” giving Ultra Mobile the opportunity to be a better partner. “We are really managing our invoicing with them and managing those business practices better than we have in the past just because we have better insight into that data.”
It’s important to know the tools you have at your disposal and the best purpose for each, he said.
“I think the best practice that we’ve done is understanding the strengths and weaknesses of the product. Because you are virtualizing from many, many different sources, you don’t have some of the functions and features like ranking that you would have if you were going against a native Data Warehouse.”
But ranking and aggregation functions can still get done at the reporting level, he said, using tools like Looker.
“It allows us to have this vast pool of data to fish from and the data is always consistent – we always know what we’re going to pull out of it. We can run the crosscheck and it’s very simple to change the hooks on the back end.”
He considers this ability to standardize data products that are coming out the system, no matter where the data comes from or what type, one of the greatest strengths of Data Virtualization.
“People go in and they look at the dashboard and they see consistency in reporting without worrying about the back end – it all just keeps humming. So really, protecting the business layer from all that chaos is what’s probably the best business benefit,” he said.
Steps Moving Forward
Now that their analysts are not spending so much time searching for their data, Burkes plans to leverage Denodo’s RESTful services piece:
“We’re starting to do our customer lifetime value retention modeling in Hadoop, and we’re going to expose those as RESTful services through the Denodo layer so that our call agents can reach in and understand the 360 components of a customer while they are on the phone.”
Burkes says the benefits of adding Data Virtualization to their toolbox have allowed Ultra Mobile the opportunity to focus on continuing to move forward at light speed:
“We’re really starting to innovate and see how we can do things very efficiently on the back end, and then how we can move into other markets. Ultra in the future [will] continue to establish ourselves as a company that really cares about our customer and is committed to providing the best value to them. Business intelligence is driving the decisions that will make it happen.”
Photo Credit: LIPING/Shutterstock.com