Analytic Methods are Best Suited for Certain Business Goals

In developing a data strategy, organizations should first determine which goals to focus on.

Match Goals with Analytic Methods

Considerations

Analytic solutions are commonly composed of one or more of the following tasks:

Match Goals with Analytic Methods (continued)

Considerations

⁴(Provost and Fawcett 21-23)

Criteria to look for:

Patterns exist that can be learned
Objective Function as target to maximize/minimize can be defined
Patterns with enough complexity to make normal decision making sub-optimal
Data needed exists or can be acquired/created
Existing data has common patterns relevant to prediction target
Problems predictive in nature⁵

Scenarios with the following characteristics are most conducive to analytic solutions:

The task is repetitive or needs to be repeated frequently
The problem or task is scalable
Patterns are constantly changing⁶

Common use cases include:

Search engines
Recommendation systems
Demand forecasting
Price optimization
Customer acquisition/retention
Fraud detection
Brand monitoring⁷

⁵(Huyen)
⁶(Huyen)
⁷(Huyen)

Match Goals with Analytic Methods (continued)

Considerations

Leaders should look for indications that analytic solutions are unsuitable. Characteristics include cases:

That are unethical
That are cost-inefficient
Where simpler solutions produce similar results
Where data cannot be acquired
Where prediction errors potentially cause catastrophic or permanent consequences
Where decisions need to be human-interpretable

Leaders should conduct a cost/benefit analysis of the potential opportunities they’ve identified and develop an evaluation framework for the expected return of the strategy. To accurately assess, they need to evaluate necessary investments in infrastructure and talent to be successful. Relevant expenses include but are not limited to:

Direct acquisition costs: Purchasing or licensing of data
Indirect acquisition costs: Opportunity costs of experimenting, including lost time, unprofitable period of testing, and data sharing incentives like loyalty card perks
Dataset Labelling
Infrastructure investments
Human capital investments
Maintenance costs