Ch. 13 Data Science & Business Strategy

Get Started. It's Free
or sign up with your email address
Rocket clouds
Ch. 13 Data Science & Business Strategy by Mind Map: Ch. 13 Data Science & Business Strategy

1. Competitive Advantage

1.1. Asset has to be valuable

1.1.1. Dell - customer personalization configuration

1.1.2. Competitor can't possess the asset to must not be able to obtain the same value

1.2. Opportunity vs Possible threat

1.2.1. Amazon vs Borders - personalized recommendations after purchasing a book

1.3. Competitive disadvantage?

2. Fundamentals of Data Science

2.1. Data Rich Firms

2.1.1. Executives, managers and investors increase their exposure to data science projects

2.1.1.1. leads to increase in opportunity

2.1.1.2. Extreme Case: Google and Amazon

2.1.1.2.1. A- "recommendations"

2.1.1.2.2. A- Cloud storage

2.1.1.2.3. G- "Prediction API"

2.2. 1980s 1990s

2.2.1. Predictive modeling

2.2.1.1. reduce cost of repairing problems in telephone networks

2.2.1.2. solve speech recognition systems

3. Competitive Advantage with Data Science

3.1. Continue to invest in new data assets

3.2. Always develop new techniques and capabilities

3.3. Achieve sustainable competitive advantage

3.3.1. Competitors can't replicate data asset or capability

3.3.2. Or at the competitors expense of replication of asset or capability

4. Formidable Historical Advantage

4.1. Can be costly to competitors

4.2. Amazon- "Dotcom Boom"

4.2.1. sold books below cost and investors rewarded company

4.2.2. could create valuable data based products (recommendations and product ratings)

4.2.2.1. VALUED BY CUSTOMERS

4.2.2.1.1. Data acquisition- catch 22 for competitors because they need customers to acquire data but need data to provide

4.2.3. Increase the cost to competitors

5. Unique Intellectual Property

5.1. Can include novel techniques for results

5.1.1. can be patented or trade secret

5.1.1.1. Competitor legally can't duplicate or will have an increased expense of doing so

5.1.1.1.1. licensing tech or developing new tech to avoid infringing patent

6. Unique Intangible Collateral Assets

6.1. Good performance

6.1.1. Effective predictive modeling

6.2. Company culture

6.2.1. Easier place for data science solutions occur when culture embraces business experimentation and rigorous supporting claims

6.2.2. Developers must understand data science

7. Superior Data Scientists

7.1. Quality and ability are a must

7.2. ACM SIGKDD holds a conference for data mining competition

7.2.1. Best scientists compete and many win year after year

7.3. Learned by practice

8. Superior Data Science Management

8.1. Managers

8.1.1. Communication

8.1.1.1. translating data science jargon into business jargon

8.1.2. Coordination

8.1.2.1. complex activities EX- integration of multiple models with constrains and costs

8.1.3. Anticipation

8.1.3.1. outcomes of data science projects through R&D

8.1.3.2. Investment guiding

8.1.3.3. see a project through

8.1.4. Understand needs

8.1.4.1. anticipate needs

8.2. Competitors can't duplicate

9. Attracting and Nurturing Data Scientists and Their Teams

9.1. The firms management must think data analytically

9.2. Firm's management must create a culture where data science and scientists thrive

9.2.1. Examples- IBM, Microsoft, Google best at hiring

9.2.2. Create an environment for nurturing data science and scientists

9.3. Cost Effective- take one+ top notch scientists as scientific advisors

9.4. Hire 3rd party to conduct

9.4.1. business analytics to data scientific consulting firms to boutique data science firms

10. Examine Data Science Case Studies

10.1. Application of data science to business problems

10.2. Ex- Perlach et al. (2013)

10.2.1. targeted consumers with online display ads, obtain adequate supply of the ideal training data

10.2.2. data was available at a lower cost from other distributions and for other target variables

10.2.3. surrogate data allowed them to operate with reduced investment

11. Evaluate Proposals for Data Science Projects

11.1. Is the business problem well specified? Does the data science solution solve the problem?

11.2. Is it clear how we would evaluate the solution?

11.3. Does the firm have data assets?

12. Example Data Mining Proposal

12.1. Target-Whiz Bang Customer Migration

13. Flaws in Big Red Proposal

13.1. Business Understanding

13.1.1. the target variable definition is imprecise

13.1.2. the formulation of the data mining problem could be better aligned with the business problem

13.2. Data Understanding/Data Preparation

13.2.1. there aren't any labeled training data. we should invest some in our budget of obtaining labels for some examples. targeting.

13.2.2. if we are worried about wasting the incentive on customers who are likely to migrate w/o we should observe a control group

13.3. Modeling

13.3.1. Linear regression is not a good choice for modeling use a classification method

13.3.1.1. tree induction, logistic regression, k-NN

13.4. Evaluation

13.4.1. the evaluation shouldnt be on the training data

13.4.2. use some cross-valuation

13.5. Deployment

13.5.1. the idea of randomly selecting customers with regression scores greater than .5 is not well considered

13.5.2. not a clear regression score of .5 responds to a probability of migration of .5 seconds

13.5.3. since our model is providing a ranking we should rank our target choose the top ranked candidates as budget allow

14. Firm's Data Science Maturity

14.1. In terms of data science capability

14.2. At the end of a maturity spectrum a firm's data science processes are complete adhoc

14.3. Medium level employs well trained data scientists who understand fundamentals, good framework and business setting

14.4. High end firms still work to improve their data science and have target customers with highest probability of leaving and even expected loss id the churn

14.5. Immature will have analytically adept employees implementing ad hoc based on intuitions (can or cannot work it's a risk)