Foundations of Business Analytics (for
GenAI Managers)
Session 1: From Business Questions to Metrics
The Turing Test
(Can Machine think?
A. M. Turing, 1950)
AI Overview
Artificial
Intelligence
(AI)
AI in the Workplace
Recent research shows that even non-routine tasks can be performed by algorithms (ML + Big
Data)
Resume screening (Cowgill 2018) Scanning CT images (Arbabshirani et al. 2018)
• Review judicial trials (Kleinberg et al. 2017)
• 47% of US employment at risk due to automation (Frey and Osborne 2013)
High Level AI Architecture
Algorithms
Sensor
Data • Knowledge- Human-
s Based
Conditioning • Unsupervised/
Machine USERS
Supervised Teaming
Source Learning
s • Transfer Learning
• Reinforced
Learning
• etc.
Modern Computing
CPUs GPUs TPUs Neuromorphic Custom Quantum
Robust AI
Explainable Metrics and Bias Verification and Security Politics, Ethics, Safety,
AI Assessment Validation (e.g Counter AI) and Training
Source: MIT Lincoln Laboratory
Today’s Objectives
1 2 3 4
See the analytics Build a KPI tree Draft a mini data Use GenAI as an
workflow for a real dictionary assistant—verify
end-to-end decision everything
Why Analytics
Now (in a
GenAI world)?
Big Data
• Collection of large datasets that cannot be
processed using traditional techniques like
regressions.
• Refers to both structured and unstructured data.
• Holds enormous promise to improve decision-
making in organizations.
Characteristics of Big Data
Variety
DIFFERENT FORMS OF DATA
Types of Analytics (Plain English)
Descriptive — What happened?
Predictive — What is likely next?
Prescriptive — What should we do?
Different management problems are best solved by different analytical approaches . . .
THREE APPROACHES TO ANALYTICS
DESCRIPTIVE PREDICTIVE PRESCRIPTIVE
Approach Business Intelligence Prediction Engines Decision Automation
What happened ? What will happen ? What should I do now ?
Role of Machine Helps me understand Supports my decision Tells me what to do
Size of Value-Creation
Opportunity
• Strategic Planning • Demand Planning • Inventory Optimization
• Initial Product Pricing • Discount / Promotion mgmt. • Price Optimization
Examples
• Scenario Planning • CRM Segmentation • Markdown optimization
• Investor Reporting • Maintenance • Risk Optimization
• Larger Size of
• Little data when compared • Quick-win opportunities
improvement prize
Rationale with problem • Relatively frequent
• High frequency of
• High uncertainty decisions & observations
decisions
• Simplified manual approach • Semi-automation
• Full automation
How do we solve business
problems with Data?
The Workflow: Decision → Metrics → Data → Analysis
→ Decision → Impact
Start with a decision, not with data
Choose a North Star KPI and supporting metrics
Define data needs (grain, owners, risks)
Analyse to inform an action; then measure impact & learn
North Star & KPI Tree
North Star: the primary definition of success
Acquisition, activation, engagement,
Supporting KPIs: levers & retention, monetisation
quality checks Operational quality (SLA, defect rates),
customer experience
Guardrails protect margin, risk, and customer trust
Example:
AirBnB
North Star →
KPI Trees
AirBnB KPI Tree
Spotify
• What is Spotify’s North Star?
• Draw the KPI Tree for Spotify
Spotify KPI Tree
Clear definition
(numerator/denominator, time
window)
What Controllable by the team; aligned to
actions
Makes a
Good Comparable over time/cohorts; stable
Metric definitions
Trustworthy (hard to game; low
leakage/bias)
Data Basics for Managers
• Rows & columns; the concept of data ‘grain’
• Keys & joins; sources & stewards/owners
• Common risks: missingness, sampling bias, data leakage
Definition of Grain: “What does one row in this dataset represent?”
Mini Data Dictionary (Template)
Name Definition Grain (row Source/Owner Risks
unit)
active_user_28 User with ≥1 user-day Product Bot traffic;
d session in last Analytics inconsistent IDs
28 days
on_time_delive % of orders order Ops/Logistics Clock sync;
ry_rate delivered by promise
promised time window drift
first_month_ch % of new users user CRM/Data Definition drift;
urn inactive in first cohort mix
30 days
Visuals That
Tell the Truth
GenAI as an
Assistant (Not
an Oracle)
Live Prompt
(Demo)
Exercise
THANK YOU!