I recently completed teaching the Spring Semester Big Data MBA: Thinking Like a Data Scientist (TLADS) class at Iowa State University. I had her 17 his MBA his sophomore year and their hard work, passion and creativity were evident throughout the semester, especially in the final presentation of the project.
There were no tests or midterms in this class and students were asked to memorize and spit out knowledge easily accessible from my books, blogs and ChatGPT. Instead, these students will learn how and where to apply data science to deliver meaningful, relevant, and ethical business outcomes in defining, synthesizing, and presenting clearly and creatively. were evaluated as a team based on their ability to work together. Yes, it was like a real consulting project, with tight deadlines, changing team relationships, and changing project scope throughout the engagement.
I made 5 teams of 3-4 students. Each team chose a company and researched that company to identify and understand the company’s key business and his initiatives. The team then applied my “think like a data scientist” methodology over the next 13 weeks to apply data and analytics to the company’s business initiatives where and how to make them meaningful and relevant. and whether they can produce ethical results. The companies and business initiatives they chose are:
- Lululemon: Doubling Male Sales While Maintaining Female Sales
- Nike: Pushing Sustainability Initiatives to Win Environmentally Conscious Customers
- Nvidia: More design wins in the game console market
- Twitter: Improve personalized content matching to increase Twitter user engagement
- John Deere: Increasing sales of hybrid/EV tractors and equipment
As we progressed throughout the semester, new ideas were explored, resulting in modifications to existing TLADS design templates and even new design templates for new TLADS. Yes, we all benefit when we are fearless in our exploration, collaboration and sharing.
Step 2 of the TLADS process seeks to understand that KPIs and indicators for key stakeholder decisions, desired outcomes, and effectiveness of outcomes are measured. The persona template in Figure 1 puts you in the shoes of your key stakeholders to better understand their work, objectives, desired outcomes, and key decisions.
shape 1: TLADS Persona Template
We’ve found that a simple decision map can complement the persona template to facilitate brainstorming of the variables and metrics that can influence key persona decisions. These “influencers” are the starter set of variables and metrics (features) leveraged in TLADS steps 5 and 6 in defining analytical scores that predict the likelihood of stakeholder behavior or performance. (Fig. 2).
shape 2: TLADS Stakeholder Determination Map
Step 5 of the TLADS methodology identifies the features used to create analytical scores that support key stakeholder decision making. Naturally, features (weather, local demographics, traffic patterns, local events, economic conditions, etc.) can be reused across multiple analytics scores. So we created Template 6.1: Use Case Feature-to-Score Mapping to help you identify and map features that you can reuse and manage in your organization’s Feature Store. (Figure 3).
shape 3: Mapping Features to Analysis Scores
The blog Features Part 1: Are Features the new Data? explained why features are more valuable economic assets than data. It would be nice to have a TLADS design template to support that hypothesis.
And now a completely new TLADS design template that I’ve been looking at for months. I have run these TLADS workshops of his for several years. And every time I reach the final step, the process feels incomplete. Especially when organizations are looking to build AI models that continuously learn and adapt. The TLADS methodology desperately needed a formal feedback process. So welcome to the new Template 8: Recommendation Presentation and Feedback template (Figure 4).
shape Four: Step 8: Recommendation presentation and feedback template
I am 99.9% sure this form will change as I get more feedback from running these TLADS workshops. And the fact that the methodology continues to learn and adapt will help ensure that I have a role to play at a time when AI technologies like ChatGPT and Google Bard are putting everyone else out of work. Means “Bill Schmarzo Full Employment Act” (Wait, do I still have to work?)
Figure 5 depicts the updated version 2 of the Thinking Like a Data Scientist (TLADS) methodology with a new step 8. This is the first major change to its methodology since it was developed many years ago. But the reality of a world where AI can create products, processes, and policies that can continuously learn and adapt required an update. It’s a brave new world.
shape Five: Think Like a Data Scientist Methodology – Updated Version 2.0
A special thanks to my students for their bold exploration of new concepts and techniques. Your energy and creativity are contagious. Can’t wait to do it again!
 Feature store is a new ML-specific data system used to centralize storage, processing, and access to frequently used features, allowing them to be reused in future machine learning model development. increase.