Generating Insights: The Gateway from Descriptive to Prescriptive Analytics


As the field of analytics matures, moving from descriptive to prescriptive analytics has become the gold standard for businesses aiming to make data-driven decisions. The linchpin that enables this transformation is the generation of actionable and valuable insights. For insights to be truly transformative, they must meet four critical criteria: they should be actionable, measurable, drive value, and offer new perspectives. In this article, we explore various methodologies for developing such insights.

The Criteria for Valuable Insights


Insights should guide decision-making processes and suggest a clear path for necessary actions.


The impact of insights should be quantifiable to understand their ROI and effectiveness.

Drives Value

The ultimate aim of any insight is to create business value, be it cost-saving, revenue-generating, or customer satisfaction-improving.


A worthwhile insight provides a new perspective or uncovers hidden patterns, setting the stage for innovative strategies.

How to Develop Insights

Attribution Approach

  1. Identify Trends and Anomalies: Begin by scanning your data for significant trends or abnormalities that stand out.
  2. Hypothesize Causes: Try to rationalize these trends by attributing them to specific actions or optimization efforts.
  3. Conduct Experiments: Design and execute experiments to test these hypotheses and determine if they are indeed responsible for the observed data patterns.

Results Approach

  1. Gather Action List: Create a list of actions or optimization efforts that have been performed.
  2. Formulate Hypotheses: Develop assumptions about the possible outcomes of these actions.
  3. Verify with Data: Compare these hypotheses with actual data trends and anomalies.
  4. Conduct Experiments: If a correlation exists, perform controlled experiments to ascertain if these actions are the real drivers of the observed outcomes.

Questioning Approach

  1. Examine Data: Look at trends and abnormalities without initially attributing them to any actions.
  2. Ask Questions: Develop questions that narrow down focus areas.
  3. Consult Experts: Use these focus areas to initiate discussions with business experts.
  4. Plan Experiments: Create potential plans or optimization efforts based on these discussions.
  5. Conduct Experiments: Before scaling, validate these plans through experimental data.


Whether it’s a attribution, results, or questioning approach, each method offers unique advantages for generating actionable and valuable insights. Importantly, the real proof of any insight’s value is validated through experimental data. By employing these approaches strategically, organizations can pinpoint which actions and optimization efforts are worth pursuing, thereby moving closer to the realm of prescriptive analytics.