International Institute for Analytics

Southwest Airlines recently launched an Enterprise Data Science Center with the objectives of expanding data science capability, deploying it broadly across the company, and creating competitive advantage. Design of the Center relied upon a series of strategic and tactical conversations with IIA Experts on analytics organization structures. Today, Southwest’s Center features disciplined delivery processes performed by data scientists in clearly defined roles who engage with the business in flexible ways. IIA’s Robert Morison collaborated with Southwest’s Justin Bundick, Director of the Enterprise Data Science Center, to capture the story.

Inquiry:

I manage our analytics team, and it has always in the marketing function. Now other business functions are interested in analytics too. How can I configure analytics to service the growing enterprise-wide needs?

Introduction In 1996, two computer science students — Larry and Sergei — were enthralled by the emerging internet. But finding anything on the undeveloped web was horribly difficult. Then came the “Aha!” discovery that academic web page citations (URLs) are a proxy for popularity. If many websites “like” the same web page, that page value is probably higher to researchers. So Larry and Sergei designed an algorithm called PageRank. It measured “link juice” — the strength between web pages. Google emerged from PageRank, web URLs and an advertising business model. This article explores the incredible value of “link juice.” Graph analysis turns the relational…

Inquiry:

What’s the right approach to pricing as an analytic function within a large retail organization? We refresh our elasticity models every few years, but is that enough?

Sanford Health is a major health system with over 49,000 employees, 187,000 health plan members, and $6.1B in annual revenue. The organization has been pursuing a growth strategy for the last two decades. Milestones include: merging with a series of regional health providers, incorporating the North Dakota Public Employees Retirement System (NDPRS) members into its health plan, and most recently the 2019 merger with Good Samaritan Society, the largest not-for-profit provider of senior housing and services in the United States.

Inquiry:

Leadership is disheartened because they don’t see enough movement with advanced analytics and artificial intelligence (AI), and we’ve been mandated to set up an AI COE. The problem is that their examples of AI are really examples of automation. There’s a gap between technology awareness and reality. How do we work around this?

Not long ago, the role of Data Scientist was what most companies wanted to discuss with me in terms of roles they needed to understand and add to their organizations. Then, the role of Data Engineer became a big topic of discussion. In the past year, there has been a massive increase of attention being paid to yet another role that is still new enough that its title hasn’t been standardized. This role is referred to by a range of names from Analytics Translator, to Analytics Catalyst, to Analytics Liaison, and more.

Inquiry:

What are use cases where machine learning (ML) delivers the best decisions and performance? Vendors are peddling their ML solutions, and because it’s sexy right now, they’re gaining traction with executives being told outside ML algorithms will outperform on every data problem.

Multi-model databases (MMDBMS) have been expanding the definition of database for several years. A multi-model database combines several data stores in one database. Those data storage services support distinct data models. Data models include relational, graph, documents, key-value, time-series, and object stores. But simply storing different kinds of data is insufficient to call it multi-model. Specialized programming services must exist for each data model. In the best MMDBMS, a single query can combine data from all data models.

Inquiry:

We’re planning to upgrade from Oracle DW to SAP Hana. For non-SAP legacy data, we’re leaning toward a traditional enterprise data warehouse (EDW), perhaps structure using Data Vault methodology. What are your thoughts on this approach?