Life Hacks for Thrilling the Customer with Your Data Science Technical Summary Products

Anne Lifton
Manager of Data Science


Anne Lifton is a Manager of Data Science at Nordstrom and in this presentation gives her own life hacks that she has come up with through her experience working in a variety of different fields that utilize machine learning and the skills that data scientists possess. Now working at Nordstrom, the goal of making the digital experience match the level of service customers expect has required a rework of the entire process of how data is seen.

“Being data-centric is a new thing and it requires a different way of thinking.”

A life hack is a strategy or technique that you can use to manage a situation in a more efficient way. When this is applied to data governance and strategy that drives many businesses today, there are oftentimes a lack of alignment within an organization can lead to opportunities for improvements. This can start by changing the management structure to focus on the theme of mathematics and statistics. There are old tools and mentalities that are good and still useful but that can be adapted within a data-centric driven goal.

(example of uncleaned data representing a binary male/female output)

The first life hack provided focuses on obtaining quality data. Efficient and successful machine learning algorithms start with a clean data set. Such a data set is able to provide a variety of different outcomes from clear and distinct variables. Getting this high-quality data is incredibly important and should have the proper time and effort devoted to ensuring this accrual is done properly and methodically. Asking your data scientists to wrangle the data in addition to developing a model from it is not always the most efficient use of time. Defining this role and responsibility should be considered carefully. 

The second life and third life hack are very closely related. The second hack is to integrate people with a variety of skill sets into a single team. Well-defined roles with specific skill sets are optimal for a business model. The third hack outlines an idea to have a Product Manager responsible for translating business needs to a Data Scientist. Defining a business problem is a task in of itself and to make sure this problem is being defined properly requires someone with the proper knowledge of the business. This is not saying that Data Scientists cannot define business problems, but it is usually not the job they were hired to do. Clearly defining whose responsibility this task should be leads to a much higher success rate of model development as it allows for specialized skill sets to be utilized more efficiently. 

The fourth life hack is a very specific one that encourages the use of tensors as opposed to the more traditional vectors or matrices. Vectors and matrices are much more commonly used as data structures due to the ability to more easily conceptually visualize and manipulate them. However, when using a tensor, an additional dimension of relationships becomes available.  This may be difficult to visualize for an individual, but for a machine learning algorithm it presents an additional opportunity to determine correlations. For many machine learning algorithms, handling tensors can be done natively without additional engineering. Neural networks can take any shape of data, including tensors, to predict results. Anne highly encourages teams to work with tensors, as it has offered many benefits and insights that may have otherwise been missed.

Anne is a great example of how the progress of how machine learning has evolved in the last decade and will continue into the future. People across industries have been implementing different models for different purposes that led to insights applicable across industries. It is people like Anne who traverse these industries and bring that experience and expertise with them that allows for this field to continue to grow and for the true power of machine learning to be recognized.

Tags   •   Retail

Recent Posts

View All Posts