Recent studies indicate that the majority (60-80%) of advanced analytics/AI projects fail, often citing “management resistance and internal politics”. However, having reviewed/supervised hundreds of such projects around the world, our experience suggests that many projects are ill-conceived and fall victim to poor analyses and pre-launch planning. This case series presents five vignettes illustrating the main “blind spots” that analytics professionals seem to overlook. These are: (i) translating the business problem and analytical problem, (ii) calculating the expected value from the project in the presence of inevitable model errors, and (iii) sequencing the steps in the project and thinking creatively about stop-loss actions. Each vignette is one page long and can be used in any course or programme on advanced analytics.
To expose learners to the pitfalls of planning analytical projects and to the main decision-analytic tools for analysing such projects
- Strategy
- Analytics
- Descriptive analytics
- Predictive analytics
- Prescriptive analytics
- Supervised learning
- Unsupervised learning
- Confusion matrix
- Model errors
- Classification
- Decision analyses
- Decision trees
- New product development
- Digital
- Q42020
