Lessons learnt from hundreds of ML Projects
Focus on 1) User. 2) Task or Decision user need to make 3) Data and information
Identify use case
Build demoable solution
Add automated deployment
Capabilities to extend to many customer
Increase utility
Use list of features
Use 2 dimensions - functional utility, customer satisfication
Determine which features are satisfying or exceeding customer need
Build strategic capability
Buy routine
For data driven capability above becomes more complicated. See slide
Convert use case to ML Use case
Determine elephant and monkey tasks
Determine whether more data, different data, different algorithm will produce results
Test
Before
Identify Research Question, Justify Why experiment is needed, Define Null Hypothesis, Do Experiment design
During
Randomization, EDA, Modeling, Analysis, Conclusion
Strategy MAP, GTM Roadmap, GTM Planning
Product Market Fit, Customer Value MAP, Brand positioning
Unique selling proposition, Product value proposition