Even the casual observer can see there is a diversity of data sources running into these companies. This poses a challenge for Dustin Spangler, VP of data and analytics at LHMSE and his primary development team of five people.
One is the elusive single customer view and customer identification. A corporation might buy several tickets for a basketball game and therefore be the customer, while the employees of that corporation would be the attendees. “Defining that customer path for each entity is slightly different, so that’s been a challenge for us. Another is that we want to understand who our customer is across multiple platforms and multiple entities, whether they go through the app or buy a ticket through a third party site.”
This is a particularly tricky issue with the Utah Jazz basketball games. LHMSE ticketing goes through Ticketmaster but fans can also buy through the NBA website which has a Utah Jazz subdomain, resulting in duplicate tracking systems.
They are currently using Teradata Vantage as their data platform and leveraging WhereScape RED as their ETL tool, as well as using Python and R. Wherescape allows his team to have much greater impact relative to the size. “It enables us to do automation and templating of our processes that makes it so that with a small team, we can work rapidly and compete with teams ten times our size in terms of volume and quality of work,” he said.
The decision to go with those tools was made three years ago. Spangler said: “We wanted to focus on the analytics over engineering. We wanted to use tools and applications that made it easy for us to just go as quickly as possible as opposed to having to write a lot of code.”
The choice of Teradata was not made lightly and came after a thorough assessment and evaluation process. His company was in the midst of a drive to move to the cloud and shortlisted three cloud vendor platforms, Google Cloud, Snowflake and Teradata, to test out.
The conditions were that they spend two weeks using each one, with one developer in charge, no early access to the data and at the end, the labour hours for the project would be reported. Essentially they wanted to see which platform would be least labour-intensive and therefore make the team most efficient.
The platforms were evaluated on: ease of use, tool integration, machine learning/artificial intelligence capabilities, maintainability, database administration, data ingestion, features and roadmap. For each criteria, the platforms were scored from one to five with Spangler joking that one meant non-existent while five meant the cyborgs were taking over.
After the evaluations, Spangler and his team concluded that all of the platforms could succeed with the right resources in terms of people, software and hardware/CPU and most of the scores given were twos, threes or fours. They also concluded that the total cost of ownership was about the speed to insights/capabilities, the analytic and reporting platform, the ability to analyse data where it lies, workload management and security.
Ultimately, they chose Teradata Vantage as its key feature is centred around time to value. Spangler also said that analytics is bigger than engineering and the other platforms required greater engineering.
Furthermore, Google and Snowflake are cloud-only so only Teradata gave them the option of using a hybrid on-prem and cloud solution. He said: “Teradata had a cloud offering that worked for us. We were still getting the performance in our testing and we felt that, in conjunction with the tools we had, we would be able to iterate and move quickly. We weren’t having to re-architect everything, we were able to leverage standard industry tools.”
Looking ahead, Spangler is looking forward to using more of the features within Vantage such as the built-in machine learning functions and the app centre which allows users to build and publish apps.
With these tools, Spangler and his team wanted to move beyond reporting what happened in the past and understanding why it happened and towards prescriptive analytics to make it easier for the businesses to plan.
An example of this is a prediction model built by team member Jeff which estimates how many people will be at a particular sports game or movie theatre screening. This allows the managers to ensure the right amount of food is available for purchase and the right number of staff are around to attend to those customers. “That for us is the next step in terms of maturity, more implementation of that prescriptive side,” said Spangler.
Dustin Spangler spoke to DataIQ at the Teradata Universe conference.
