Please enable JavaScript.
Coggle requires JavaScript to display documents.
Provost Chapter 13 Monique Bromfield (Competitive Advantage with Data…
Provost Chapter 13
Monique Bromfield
Competitive Advantage with Data Science
The asset (data) should be valuable in the context
of company's strategy
Always stay ahead of the competition
Aways be developing new techniques and capabilities
Always be investing in new data assets
Historical data is always an advantage (may not be able to recreate)
Creates high switching costs
Unique Intellectual property (data mining techniques = patented or a simple trade secret)
Superior data sciencetists
Have good management of data science team
Understand needs of business
Communicate well between different business aspects
Anticipate outcomes of projects
Coordinate technically complex activities
Attracting and nurture data scientists and their teams better than competition
Examine data science case studies
Accept creative ideas - can lead to new ways of collecting data / different data
Evaluate Data Science Proposals
Flaws in proposals
Business Understanding
Target variable definition is imprecise
The data mining problem could be better aligned with the business problem
Data Understanding/Data Preparation
Observe control groups
Labeled training data
Modeling
Linear regression is not a good choice for modeling a categorical target variable
Evaluation
Shouldn't be on training data
Will there be any domain-knowledge validation of the model?
Deployment
Randomly selecting customers with regression scores greater than 0.5 (arbitrary) is not advised.
Questions to ask
Does firm have the data assets?
Will we see evidence of success?
Is the problem well specified?
How will we evaluate solution?
Maturity of Data Firm
How systematic & well founded the processes used to guide the firm’s data science projects are
Medium Level - employs mature data scientists & business managers to solve data science problems
High End - work to improve their processes & executives constantly challenge team