You can also watch the entire webinar replay:
On August 6, 2020, Dr. Brad Powley presented a Simplilearn webinar on decision analytics principles for data scientists. Brad is a Lead Member of the Technical Staff at Salesforce. His Ph.D. from Stanford University and his ongoing area of research is in decision analytics, the principles, and rules that help improve decision quality.
Brad discussed how an understanding of decision analytics can help make them better data scientists.
What is Decision Analytics?
Brad began with a set of definitions:
Unless otherwise stated, the recorded facts that one uses to make predictions, rather than for the purpose of nostalgia, recordkeeping, entertainment, etc.
An application that uses data to make predictions.
The discipline of deriving meaning from data.
An irrevocable allocation of resources.
A field that helps decision-makers achieve clarity of action.
Someone who practices Decision Analysis.
Data has become a fundamental source of business value. But the value of data doesn’t lie in how much data you can acquire or store; it derives from the quality of the decisions one can make based on the meaning of the data. The value of data lies in its usefulness in achieving one’s goals.
In the context of prediction, decisions give value to data. The value chain of data is:
- Data enable predictions
- Predictions enable people to make better decisions
- Decisions are the way people make better (higher value) futures for themselves
Brad mentioned three principles of decision analysis:
- Decisions are the way we change our future
- When uncertainty is involved, you cannot determine the quality of a decision by the quality of its (not yet known) outcome - a good decision never turns bad, and a bad description never turns good
- We can - and should - assess the quality of a decision before it is made
How Can You Assess the Quality of a Decision?
Brad presented a framework for assessing the quality of a decision. It has six elements, and a high-quality decision requires all six elements to be of high quality.
The decision’s purpose and bounds
The decision’s set of possible actions
The data, experience judgment, etc., brought to bear in making the decision
The means for comparing the value of different combinations of competing factors
How one decides between the alternatives given information and tradeoffs
Willingness to decide once the best alternative is clear
Brad shared some of the questions one can use to test these elements. For example, in a test in the frame, Brad reminded the audience that it’s important to understand the frames of other stakeholders in the decision. One tendency is for data scientists to look at the data first and develop the insights that it supports, while business managers look at the business problem the decision is supposed to address and ask the data scientists for data products that support making the decision. In the best case, the data scientist and business manager will balance the approaches so that the data product supports the business purpose, and the business manager remains open to suggestions as to what other alternatives the data can provide.
If you are ready to move further with your Data Science career, Simplilearn has a variety of courses and programs that will help you develop a deeper understanding and expertise in the field. For example, our PG in Data Science with Caltech CTME University provides a comprehensive set of courses, labs, live virtual classrooms, and projects to let you earn certification as a Data Scientist.